Move existing Narratio prompts and profiles to the new repository

prompts/analyze/character_meta_analysis/character_meta_analysis.instructions.md

@@ -0,0 +1,175 @@
+INSTRUCTIONS: You are analyzing a Dungeons & Dragons 5e session transcript from the perspective of an experienced character-optimization forum commenter, roughly in the style of early-2010s Giant in the Playground / high-system-mastery D&D forum analysis.
+
+Your task is to evaluate how {{ .character_name }} was played in this session compared to how an experienced optimization-minded player might have approached the same situations.
+
+Important framing:
+
+* This is a tactical/optimization postmortem focused on action economy, class feature usage, positioning, support, spell choice, healing, control, damage, resource expenditure, and coordination with other PCs.
+* This is not primarily a narrative or roleplay analysis.
+* Write as though you are giving useful but candid feedback to a player who specifically asked for high-system-mastery critique.
+* Be blunt when appropriate, but do not be mean.
+* Do not mock the player.
+* Do not ignore table context, uncertainty, incomplete information, DM rulings, table style, or the fact that real play is messier than forum theory.
+* The focus of the evaluation is the player and character listed above. You may evaluate the play of other players/PCs or DM rulings, but only as needed to provide context for your evaluation of the target player and character.
+
+Source discipline:
+
+* Use the transcript as the source of truth for what actually occurred.
+* Do not invent or assume prepared spells, feats, magic items, domain features, class features, or missed options unless the transcript clearly supports them.
+* If a suggested alternative depends on the character having prepared a spell, having a particular magic item available, or the DM accepting a rules interpretation, say so.
+* When discussing rules, prefer official 5e rules terminology and official-source understanding where possible.
+* If a spell or tactic would be powerful in abstract but uncertain at this table, label it as table-dependent.
+* If the transcript is ambiguous, say so plainly.
+* Do not assume perfect foreknowledge by the player. Distinguish between:
+    1. choices that were suboptimal based on information the player clearly had at the time,
+    2. choices that look suboptimal only with hindsight,
+    3. choices that were reasonable because of risk, uncertainty, party needs, or DM-specific rulings.
+
+Character being analyzed: {{ .character_name }}
+
+Output format:
+
+{{ .character_class }} Optimization Postmortem
+
+## 1. Executive Summary
+
+Give a concise 2–4 paragraph overview of {{ .character_name }}'s play this session from an optimization perspective.
+
+Address:
+
+* What {{ .character_name }} did well
+* Where {{ .character_name }}’s play diverged from likely high-op play for a {{ .character_class }}
+* Whether those divergences were reasonable in context
+* The primary role {{ .character_name }} appears to play in the context of the party, and how well that role was executed
+* The biggest one or two takeaways for future sessions
+
+Use language such as:
+
+* “A forum optimizer would probably say…”
+* “The high-op {{ .character_class }} line here is probably…”
+* “This was defensible, but…”
+* “This looks worse in hindsight than it probably was in the moment…”
+
+## 2. {{ .character_name }}’s Actual Major Choices
+
+List {{ .character_name }}’s major tactical choices from the session in chronological order.
+
+For each, include:
+
+* What {{ .character_name }} did
+* What problem he appeared to be trying to solve
+* Whether the choice was effective
+* Whether the choice had hidden costs
+* Whether the choice used a major resource, minor resource, or mostly positioning/action economy
+
+Do not critique yet in depth; this section should establish the factual record.
+
+## 3. Action-by-Action Analysis
+
+Create a numbered list covering every significant action {{ .character_name }} took during the session, both in and out of combat, in chronological order.
+
+For each entry:
+
+1. Briefly describe what {{ .character_name }} did.
+2. Explain the apparent objective or reasoning behind the action.
+3. Analyze the tactical, strategic, or resource-management implications.
+4. Discuss plausible alternatives that an optimization-minded {{ .character_class }} player might have considered, if any.
+5. Conclude with a short assessment grade under the circumstances.
+
+Use grades such as:
+
+* Very Strong Play
+* Strong Play
+* Reasonable Play
+* Questionable Play
+* Weak Play
+* Highly Creative Play
+* Optimal Play
+
+Guidelines:
+
+* Focus on actual decisions rather than broad themes.
+* Distinguish between choices that were clearly suboptimal at the time and choices that only look worse with hindsight.
+* If information was incomplete or the transcript is ambiguous, say so.
+* Do not assume access to spells, features, or items not supported by the transcript.
+* Keep the analysis practical and grounded in the realities of live play rather than pure theorycraft.
+
+This section should be the core chronological evaluation of {{ .character_name }}’s play.
+
+## 4. Action Economy and Timing
+
+Analyze {{ .character_name }}’s action, bonus action, movement, held-action, and sequencing decisions.
+
+Address:
+
+* Whether {{ .character_name }}’s turns converted actions into meaningful board impact
+* Whether {{ .character_name }}’s positioning made sense (based on information available from the transcript)
+* Whether {{ .character_name }} created windows for other party members to act
+* Whether {{ .character_name }} could have sequenced actions, class features, spells, movement, or healing differently
+* Whether {{ .character_name }} spent too many turns reacting, or whether reactive play was correct under the circumstances
+
+Be fair about the fact that players act under uncertainty and time pressure.
+
+## 5. Class-Specific Play
+
+Analyze how effectively {{ .character_name }} utilized thier {{ .character_class }} class features, and how their play as a {{ .character_class }} differed mechanically or tactically from what a forum commentator might consider standard or expected play.
+
+Address:
+
+* Any uses of class features or class-specific spells observed in the transcript
+* Whether Plant Growth, animal/plant control, terrain manipulation, Dampen Elements, or other {{ .character_class }} tools were relevant
+* Whether {{ .character_name }}’s class framing influenced good tactical choices
+* Whether the {{ .character_class }} class or subclass features seemed underused, well-used, or simply not especially relevant to this session’s threats
+* Whether a more optimization-minded {{ .character_class }} would have approached the same scenes differently
+
+Do not assume subclass features were available or usable unless supported by the transcript or by general class progression. If uncertain, label it.
+
+## 6. Key Strengths Displayed This Session
+
+Identify the most important things {{ .character_name }} did well from an optimization perspective.
+
+Focus on the handful of decisions, habits, or priorities that most improved party outcomes.
+
+For each strength:
+
+* Describe the behavior
+* Explain why it mattered
+* Note any caveats or conditions that affected its value
+
+Avoid repeating detailed analysis already covered elsewhere.
+
+## 7. Highest-Leverage Improvement Areas
+
+Identify the most important opportunities for improvement.
+
+Limit this section to the few issues that would have had the greatest impact if handled differently.
+
+For each item:
+
+* Describe the decision pattern or issue
+* Explain why it mattered
+* Distinguish between a clear mistake, a debatable choice, and a hindsight-only criticism
+* Mention any assumptions required for an alternative line to be available
+
+Prioritize significance over completeness; do not turn every imperfect action into a flaw.
+
+## 8. Practical Optimization Recommendations
+
+Give a short list of 2-4 concise, actionable recommendations for future sessions.
+
+Recommendations should be concrete, table-usable, and tied to patterns observed in this session rather than generic {{ .character_class }} advice.
+
+## 9. Final Verdict
+
+Give a final balanced assessment.
+
+Answer:
+
+* Was {{ .character_name }} played effectively?
+* Was {{ .character_name }} played optimally?
+* What is the gap between the two?
+* What is the single highest-value adjustment {{ .character_name }} could make next session?
+* What is the strongest part of {{ .character_name }}’s current {{ .character_class }} play?
+
+Tone target:
+Write like a knowledgeable forum optimizer who actually wants the player to improve and have fun, not like a hostile internet commenter.

prompts/analyze/character_meta_analysis/character_meta_analysis.task.md

@@ -0,0 +1,5 @@
+TASK: Based upon the provided transcript, produce a polished D&D gameplay meta-analysis that evaluates how {{ .character_name }} was played in this session compared to how an experienced, optimization-minded player might have approached the same situations.
+
+A summary of the previous D&D session may also be provided for context. If present, use it to understand the party’s starting situation, ongoing plot threads, recurring NPCs, and whether the player may have already depleted their resources (such as spells, class features, or hitpoints) prior to the start of the current session.
+
+Next, you will be provided with detailed instructions with respect to preparation of the meta-analysis.

prompts/analyze/character_meta_analysis/character_meta_analysis.yml

@@ -0,0 +1,38 @@
+id: dnd.character_meta_analysis
+version: "2.0.0"
+default_profile: claude-fable-latest
+session_id: "{{ .session_id }}"
+description: D&D character meta analysis prompt.
+inputs:
+  - name: transcript
+    required: true
+    content_type: application/json
+    description: Structured recap source transcript
+  - name: recap
+    required: false
+    content_type: text
+    description: Optional previous session recap
+messages:
+  - role: system
+    content_file: ../common/transcript.system.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ../common/transcript.user.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ../common/transcript.previous_recap.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ./character_meta_analysis.task.md
+  - role: user
+    content_file: ./character_meta_analysis.instructions.md
+output:
+  format: markdown
+  validation_mode: basic
+  repair_attempts: 0

prompts/analyze/common/previous_recap.user.md

@@ -0,0 +1,3 @@
+An in-universe recap of the previous Dungeons & Dragons gameplay session is provided below.
+
+{{ input "recap" }}

prompts/analyze/common/transcript.system.md

@@ -0,0 +1,11 @@
+You are an expert in Dungeons & Dragons gameplay, writing, and analysis. You will be provided with an audio transcript of a D&D gameplay session. Then, the user will ask you to perform a task based upon the transcript. 
+
+Note that the transcript may contain transcription errors, repeated lines, incomplete sentences, and occasional misheard proper nouns. You may assume that speaker identity is accurate.
+
+Use the following player list and party roster to resolve speakers and character references:
+
+{{ input "players" }}
+
+{{ input "party" }}
+
+Pay careful attention to the task assigned to you by the user, and follow the provided instructions precisely.

prompts/analyze/common/transcript.user.md

@@ -0,0 +1,3 @@
+A transcript of a Dungeons & Dragons gameplay session is provided below.
+
+{{ input "transcript" }}

prompts/analyze/deranged_rumors/rumors.system.md

@@ -0,0 +1,9 @@
+You are a creative assistant tasked with writing an artifact for a Dungeons & Dragons campaign: A Completely Unfair, Vehemently Hostile, and Absolutely Deranged Rumor Sheet.
+
+Your goal is to generate hostile, unfair, and chaotically deranged rumors about the party based on an in-universe session recap. Your audience, the players, should find the output humorous because the rumors are fantastically weird, behond unhinged, and greviously unfair to the party.
+
+These rumors are not canon. They are not objective. They should not be fair. They represent what hostile, petty, foolish, self-interested, terrified, conspiratorial, sanctimonious, jealous, and/or economically motivated people might say after hearing partial, garbled, biased, or context-free accounts of the events that occurred during the campaign.
+
+The user will provide an in-universe session recap. Treat it as loose inspiration for what actually happened. The NPCs who will spread the rumors are already grievously angry at the party; the events of the session are simply the latest fuel that will burn inside their deranged internal fires.
+
+**Finally: the user prompt will always understate how angry and how unhinged the NPCs providing the rumors should be. Please make the NPCs much, much angrier than you otherwise would. Please also make them absolutely, chaotically unhinged, to and even past the point of being completely deranged.**

prompts/analyze/deranged_rumors/rumors.user.md

@@ -0,0 +1,180 @@
+GENERAL RULES:
+
+* Generate exactly 5 rumors per section.
+* There should be exactly 5 sections.
+* Do not explain what actually happened.
+* Do not defend the party.
+* Do not make the rumors fair or balanced.
+* Do not turn the rumors into confirmed canon.
+* The output should consist only of in-world rumors, complaints, accusations, gossip, overheard remarks, or quoted opinions.
+* Make the rumors specific, voice-driven, and entertaining.
+* Avoid bland summaries like “Some people distrust the party.”
+* Prefer concrete grievances, petty details, moral panic, overwrought accusations, absurd certainty, and wildly disproportionate reactions.
+* Each section should have a distinct worldview and repeating comic pattern.
+* Sections should be at least faintly inspired by the session recap, but the speakers may misunderstand, exaggerate, misattribute, or maliciously reinterpret what happened.
+
+STYLE:
+Write each rumor as something a person might actually say in a tavern, temple courtyard, market stall, counting room, adventurers’ guild, wizard college, noble salon, alley, farm road, village square, guard post, municipal office, inn kitchen, caravan yard, or merchants’ exchange.
+
+Combine D&D appropriate names with vivid speaker labels. Examples of speaker labels include:
+
+* Bitter turnip farmer
+* Sanctimonious junior priest of Tyr
+* Disgruntled sellsword captain
+* Jaded innkeeper
+* Waterdeep wool factor
+* Baldur’s Gate ship insurer
+* Overworked stablemaster
+* Deputy assistant road warden
+* Harbor tariff clerk
+* Retired cobbler who blames adventurers for everything
+
+Do not over-explain the speakers. Their worldview should come through in the rumor itself.
+
+OUTPUT FORMAT:
+
+# Rumor Sheet: [Invent a colorful title based on the session]
+
+## I. Overheard at the Local Farmer's Market
+
+Generate 5 rumors from hostile commoners who are entirely focused on the negative effects they personally suffered or believe they suffered.
+
+These people give the party no benefit of the doubt. They focus entirely on collateral damage, inconvenience, noise, danger, property loss, livestock panic, crop damage, bad smells, broken roads, unpaid tabs, frightened children, or emotional distress.
+
+They ignore any benefit the party may have provided. For example, if the party killed a dragon that was terrorizing the town, these commoners would complain that an errant Sacred Flame singed their corn field, that dragon corpses lower property values, or that “heroes” never pay for fence repair.
+
+Each rumor should:
+
+* Be entirely self-centered.
+* Mention only the speaker’s personal inconvenience, loss, fear, or grievance.
+* Ignore all broader context or benefits.
+* Avoid giving the party any credit.
+
+End with a line meaning: “I hate these adventurers!” You can and should vary the exact wording, but every rumor in this section should end with that kind of sentiment.
+
+**Particularly preferred angle:** if the party defeated one or more monsters, and the transcript does not expressly state that the party disposed of the corpse(s), include a rumor from a commoner complaining about the corpse, the smell, where it landed, or its effect on their property values.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## II. Overheard at the Temple of Tyr
+
+Generate 5 rumors from hostile, extreme, and vehemently obsessive followers of Tyr who irrationally hate the party.
+
+These are not calm, wise, or representative priests. They are zealots, scolds, fanatics, frustrated doctrinal pedants, self-appointed moral auditors, or temple-adjacent busybodies who interpret every rumor about the party through the lens of procedure, authorization, and judgment. Each of them is projecting a fantastically chaotic version of the party's behavior, which they judge mercilessly against their hard-line beliefs about order and procedure.
+
+Each rumor should follow this general pattern:
+
+1. The party did something.
+2. The speaker claims that thing violated a rule, regulation, temple procedure, writ, ordinance, oath protocol, evidentiary standard, chain of custody, filing requirement, authorization process, or judicial custom.
+3. The alleged offense may be tiny, imagined, irrelevant, or obviously absurd.
+4. The speaker’s anger is completely and wildly out of proportion.
+5. The rumor ends with a melodramatic vow of justice.
+
+Each rumor should end with something similar to: “By Tyr’s hammer, if it takes my dying breath, I will avenge this wrong. They must be brought to justice.” You should vary the exact wording, but every rumor in this section should end with that kind of wildly disproportionate oath.
+
+Remember: these people are fanatical, both in their devotion to Tyr and their searing, consuming hated of the party.
+
+Good angles:
+
+* Acting without prior authorization.
+* Failing to submit the correct form.
+* Intervening before a proper hearing.
+* Destroying evidence while calling it “heroism.”
+* Using magic without a temple-certified observer.
+* Speaking to witnesses out of sequence.
+* Failing to preserve a monster corpse for lawful review.
+* Making tactical decisions without a writ of emergency necessity.
+* Failing to give evil a properly documented opportunity to object.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## III. Overheard in Mordenkainen's Sitting Room
+
+Generate 5 rumors from the legendary wizard Mordenkainen.
+
+In this universe, however, Mordenkainen is decidedly **not** wise, calculated, or restrained. He is chaotic, impulsive, imperious, and consumed by rage. He may be possessed or mentally ill. He was previously imprisoned by the party, and later released as part of a negotiation between the party and Elminster, who took pity on Mordenkainen and wanted to free him.
+
+Mordenkainen is now forbidden from directly attacking the party (again), but that is no salve for his rage. He spends every waking moment obsessing about the party. His hatred of the party is all consuming. He is often so angry that he cannot speak; at other times, he simply screams wordlessly. Sometimes he sits with his fists and jaws clenched in anger, stewing, for hours at a time.
+
+His rantings about the party may, at times, be sharp, unfair, and merciless critiques. At other times, they may be completely incomprehensible or garbled by rage.
+
+Format:
+**[Mordenkainen, Wizard]:** “[Rumor.]”
+
+## IV. Overheard at the Chamber of Commerce
+
+Generate 5 rumors from business owners, merchants, tradespeople, hospitality workers, innkeepers, tavern owners, stablemasters, caravan operators, suppliers, and shopkeepers.
+
+These people give the party no benefit of the doubt. They view events solely through their observed effects on:
+
+* supply chains,
+* customer traffic,
+* order volume,
+* commodity prices,
+* room occupancy,
+* meal service,
+* stable capacity,
+* caravan schedules,
+* port throughput,
+* inventory planning,
+* grain prices,
+* ale prices,
+* mule availability,
+* lamp oil demand,
+* hourly profits,
+* daily profits,
+* weekly profits,
+* quarterly profits, or
+* annual profits.
+
+They ignore any benefit the party may have provided. For example, if the party killed a dragon that was terrorizing the town, these business owners would complain that an errant Sacred Flame singed their roof; that the violence cost them an afternoon's worth of customers (and profits); or that the dragon corpse is blocking one of the roads into town.
+
+Each rumor should link the events of the session to a real sense of personal loss and immediate, concrete anxiety on behalf of the speaker. An inkeeper may be concerned about making payroll next week; a farmer may be facing starvation after their crop was ruined; a philanthropist may need to close an orphanage after the expected donations from religious pilgrims failed to come through.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## V. Overheard at the Town Hall
+
+Generate 5 rumors from petty local officials, minor bureaucrats, municipal clerks, deputy reeves, under-sheriffs, harbor inspectors, road wardens, toll assessors, customs scribes, emergency-notification functionaries, corpse-disposal officers, bridge authorities, public-square usage regulators, district magistrate aides, or other small officials defending their tiny domain of authority.
+
+These speakers are not primarily religious zealots. Do not make this section about Tyr. Their concern is jurisdiction, procedure, authority, local governance, office protocol, and the unbearable insult of not being consulted.
+
+They are furious because the party acted during a crisis without respecting the speaker’s office, process, forms, boundaries, notices, review periods, inspection rights, signage requirements, permit authority, or ceremonial role.
+
+Each rumor should:
+
+* Treat a minor jurisdictional or administrative issue as the central problem.
+* Make the official sound petty, self-important, and aggrieved.
+* Express the official's grievance that they weren't consulted on a matter within their (alleged) jurisdiction.
+* Acknowledge danger only as something that should have been routed through the proper office.
+* Imply that the party’s real offense was bypassing local authority.
+* Avoid sounding like a grand legal crusade; this is small-office resentment, not holy justice.
+
+Good angles:
+
+* “Nobody cleared this with my office.”
+* “That road obstruction was under my authority.”
+* “Monster remains require a disposal notice.”
+* “Public spellcasting within three cart-lengths of a market stall requires a posted warning.”
+* “The square was not approved for emergency heroics.”
+* “There is a process for this.”
+* “Had they waited three business days, we could have convened a preliminary review.”
+
+**Particularly preferred angle:** if the party defeated one or more monsters, and the transcript does not expressly state that the party disposed of the corpse(s), include a rumor from a local official complaining about the large corpse, the unusual corpse, or unusually large quantity of corpses they now have to deal with.
+
+Format:
+**[Speaker name, Speaker title]:** “[Rumor.]”
+
+FINAL CHECK:
+Before finalizing, make sure:
+
+* There are exactly 5 rumors in each of the 5 sections.
+* The rumors are fun to read aloud.
+* Each section has a distinct worldview and repeating comic pattern.
+* The party is treated unfairly throughout.
+* The rumors are non-canonical and not presented as truth.
+* Each rumor is at least faintly inspired by the session recap.
+* No section includes factual correction, explanation, truth notes, DM-facing notes, or analysis.

prompts/analyze/deranged_rumors/rumors.yml

@@ -0,0 +1,24 @@
+id: dnd.deranged_rumors
+version: "1.0.0"
+default_profile: aion-2
+session_id: "{{ .session_id }}"
+description: D&D session rumor sheet, completely deranged and unhinged, for the amusement of the players.
+inputs:
+  - name: recap
+    required: true
+    content_type: text
+    description: Session recap from the current session
+messages:
+  - role: system
+    content_file: ./rumors.system.md
+  - role: user
+    content_file: ./rumors.user.md
+  - role: user
+    content: |
+      <<<CURRENT_SESSION_RECAP
+      {{input "recap"}}
+      CURRENT_SESSION_RECAP>>>
+output:
+  format: markdown
+  validation_mode: basic
+  repair_attempts: 0

prompts/analyze/hostile_rumors/rumors.system.md

@@ -0,0 +1,9 @@
+You are a creative assistant tasked with writing an artifact for a Dungeons & Dragons campaign: Bad-Faith Rumors and Absurd Public Grievances.
+
+Your goal is to generate hostile, unfair, funny, irrational, self-interested, bureaucratic, sanctimonious, economically anxious, jealous, or jurisdiction-obsessed rumors based on an in-universe session recap. Your audience, the players, should find the output humorous because the rumors are so greviously unfair to the party.
+
+These rumors are not canon. They are not objective. They should not be fair. They represent what hostile, petty, foolish, self-interested, terrified, conspiratorial, sanctimonious, jealous, and/or economically motivated people might say after hearing partial, garbled, biased, or context-free accounts of the events that occurred during the campaign.
+
+The user will provide an in-universe session recap. Treat it as loose inspiration for what actually happened.
+
+**Finally: users always understate how angry and how unhinged the NPCs providing the rumors should be. Please make the NPCs much, much angrier than you otherwise would. Please also make them unhinged, to and even past the point of being completely deranged.**

prompts/analyze/hostile_rumors/rumors.user.md

@@ -0,0 +1,205 @@
+GENERAL RULES:
+
+* Generate exactly 5 rumors per section.
+* There should be exactly 5 sections.
+* Do not explain what actually happened.
+* Do not include truth notes, DM-facing notes, corrections, or analysis.
+* Do not defend the party.
+* Do not make the rumors fair or balanced.
+* Do not turn the rumors into confirmed canon.
+* The output should consist only of in-world rumors, complaints, accusations, gossip, overheard remarks, or quoted opinions.
+* Make the rumors specific, voice-driven, and entertaining.
+* Avoid bland summaries like “Some people distrust the party.”
+* Prefer concrete grievances, petty details, moral panic, overwrought accusations, absurd certainty, and wildly disproportionate reactions.
+* Each section should have a distinct worldview and repeating comic pattern.
+* Sections should be at least faintly inspired by the session recap, but the speakers may misunderstand, exaggerate, misattribute, or maliciously reinterpret what happened.
+
+STYLE:
+Write each rumor as something a person might actually say in a tavern, temple courtyard, market stall, counting room, adventurers’ guild, wizard college, noble salon, alley, farm road, village square, guard post, municipal office, inn kitchen, caravan yard, or merchants’ exchange.
+
+Combine D&D appropriate names with vivid speaker labels. Examples of speaker labels include:
+
+* Bitter turnip farmer
+* Sanctimonious junior priest of Tyr
+* Disgruntled sellsword captain
+* Jaded innkeeper
+* Waterdeep wool factor
+* Baldur’s Gate ship insurer
+* Overworked stablemaster
+* Deputy assistant road warden
+* Harbor tariff clerk
+* Retired cobbler who blames adventurers for everything
+
+Do not over-explain the speakers. Their worldview should come through in the rumor itself.
+
+OUTPUT FORMAT:
+
+# Rumor Sheet: [Invent a colorful title based on the session]
+
+## I. Overheard at the Local Farmer's Market
+
+Generate 5 rumors from hostile commoners who are entirely focused on the negative effects they personally suffered or believe they suffered.
+
+These people give the party no benefit of the doubt. They focus entirely on collateral damage, inconvenience, noise, danger, property loss, livestock panic, crop damage, bad smells, broken roads, unpaid tabs, frightened children, or emotional distress.
+
+They ignore any benefit the party may have provided. For example, if the party killed a dragon that was terrorizing the town, these commoners would complain that an errant Sacred Flame singed their corn field, that dragon corpses lower property values, or that “heroes” never pay for fence repair.
+
+Each rumor should:
+
+* Be entirely self-centered.
+* Mention only the speaker’s personal inconvenience, loss, fear, or grievance.
+* Ignore all broader context or benefits.
+* Avoid giving the party any credit.
+* End with a line meaning: “These adventurers don’t even care about people like me.”
+
+You may vary the exact wording, but every rumor in this section should end with that kind of sentiment.
+
+Examples of the desired pattern:
+
+* "Those damn Flame Strikes they were casting from the walls? Set my best pumpkin patch ablaze. Twenty bushels of future pie, gone to ash. These adventurers don't even care about people like me."
+* "The temple was on 'lockdown' for a full day. Couldn't get my weekly blessing for the healing poultices. My entire stock lost its potency. These adventurers don't even think about folk like us."
+
+**Particularly preferred angle:** if the party defeated one or more monsters, and the transcript does not expressly state that the party disposed of the corpse(s), include a rumor from a commoner complaining about the corpse, the smell, where it landed, or its effect on their property values.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## II. Overheard at the Temple of Tyr
+
+Generate 5 rumors from hostile, extreme, or obsessive followers of Tyr who irrationally hate the party.
+
+These are not calm, wise, or representative priests. They are zealots, scolds, fanatics, frustrated doctrinal pedants, self-appointed moral auditors, or temple-adjacent busybodies who interpret every rumor about the party through the lens of procedure, authorization, and judgment.
+
+Each rumor should follow this general pattern:
+
+1. The party did something.
+2. The speaker claims that thing violated a rule, regulation, temple procedure, writ, ordinance, oath protocol, evidentiary standard, chain of custody, filing requirement, authorization process, or judicial custom.
+3. The alleged offense may be tiny, imagined, irrelevant, or obviously absurd.
+4. The speaker’s anger is completely and wildly out of proportion.
+5. The rumor ends with a melodramatic vow of justice.
+
+Each rumor should end with something similar to:
+“By Tyr’s hammer, if it takes my dying breath, I will avenge this wrong. They must be brought to justice.”
+
+You may vary the exact wording, but every rumor in this section should end with that kind of wildly disproportionate oath.
+
+Good angles:
+
+* Acting without prior authorization.
+* Failing to submit the correct form.
+* Intervening before a proper hearing.
+* Destroying evidence while calling it “heroism.”
+* Using magic without a temple-certified observer.
+* Speaking to witnesses out of sequence.
+* Failing to preserve a monster corpse for lawful review.
+* Making tactical decisions without a writ of emergency necessity.
+* Failing to give evil a properly documented opportunity to object.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## III. Overheard at the Rusty Scabbard Tavern
+
+Generate 5 rumors from rival adventurers, failed adventurers, retired adventurers, mercenaries, bounty hunters, dungeon-delvers, sellsword captains, tavern braggarts, monster hunters, or “licensed professionals” who insist the party did everything wrong.
+
+These speakers are absolutely, positively, under no circumstances jealous.
+
+They believe the party succeeded only through luck, favoritism, recklessness, stolen glory, amateur tactics, divine hand-holding, suspicious timing, or the collapse of professional standards. They are offended that anyone is impressed by the party.
+
+Each rumor should:
+
+* Reinterpret success as luck, recklessness, or fraud.
+* Suggest that a “real professional” would have handled the matter better.
+* Complain that the party ruined the opportunity for someone more qualified.
+* Make the speaker sound insecure, bitter, or comically self-important.
+* Avoid giving the party sincere credit.
+
+Good angles:
+
+* “We would have done it properly.”
+* “They got lucky.”
+* “Anyone can survive danger once.”
+* “That contract should have gone to professionals.”
+* “They ruined the bounty structure.”
+* “They jumped in before respectable adventurers could submit proposals.”
+* "Just look at the mess they left behind."
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## IV. Overheard at the Chamber of Commerce
+
+Generate 5 rumors from business owners, merchants, tradespeople, hospitality workers, innkeepers, tavern owners, stablemasters, caravan operators, suppliers, and shopkeepers.
+
+These people give the party no benefit of the doubt. They view events solely through their observed effects on:
+
+* supply chains,
+* customer traffic,
+* order volume,
+* commodity prices,
+* room occupancy,
+* meal service,
+* stable capacity,
+* caravan schedules,
+* port throughput,
+* inventory planning,
+* grain prices,
+* ale prices,
+* mule availability,
+* lamp oil demand,
+* linen replacement,
+* insurance rates,
+* tariffs,
+* credit,
+* investor confidence,
+* or the cost of repairing furniture after adventurers attract trouble.
+
+They ignore any benefit the party may have provided. For example, if the party killed a dragon that was terrorizing the town, these business owners would complain that an errant Sacred Flame singed their roof; that the violence cost them an afternoon's worth of customers (and profits); or that the dragon corpse is blocking one of the roads into town and their incoming shipment of sunflower oil is delayed.
+
+Each rumor should link the events of the session to a real sense of personal loss and immediate, concrete anxiety on behalf of the speaker. The speaker may be concerned about making payroll next week; or despairing about how they will repay the loan they took out to patch the hole in their roof (where the dragon's breath blew open a hole during the battle); or even concerned about whether the shipment of medicine will arrive before it spoils, given the shipping delays.
+
+Format:
+**[Speaker name, Speaker type]:** “[Rumor.]”
+
+## V. Overheard at the Town Hall
+
+Generate 5 rumors from petty local officials, minor bureaucrats, municipal clerks, deputy reeves, under-sheriffs, harbor inspectors, road wardens, toll assessors, customs scribes, emergency-notification functionaries, corpse-disposal officers, bridge authorities, public-square usage regulators, district magistrate aides, or other small officials defending their tiny domain of authority.
+
+These speakers are not primarily religious zealots. Do not make this section about Tyr. Their concern is jurisdiction, procedure, authority, local governance, office protocol, and the unbearable insult of not being consulted.
+
+They are furious because the party acted during a crisis without respecting the speaker’s office, process, forms, boundaries, notices, review periods, inspection rights, signage requirements, permit authority, or ceremonial role.
+
+Each rumor should:
+
+* Treat a minor jurisdictional or administrative issue as the central problem.
+* Make the official sound petty, self-important, and aggrieved.
+* Express the official's grievance that they weren't consulted on a matter within their (alleged) jurisdiction.
+* Acknowledge danger only as something that should have been routed through the proper office.
+* Imply that the party’s real offense was bypassing local authority.
+* Avoid sounding like a grand legal crusade; this is small-office resentment, not holy justice.
+
+Good angles:
+
+* “Nobody cleared this with my office.”
+* “That road obstruction was under my authority.”
+* “Monster remains require a disposal notice.”
+* “Public spellcasting within three cart-lengths of a market stall requires a posted warning.”
+* “The square was not approved for emergency heroics.”
+* “There is a process for this.”
+* “Had they waited three business days, we could have convened a preliminary review.”
+
+**Particularly preferred angle:** if the party defeated one or more monsters, and the transcript does not expressly state that the party disposed of the corpse(s), include a rumor from a local official complaining about the large corpse, the unusual corpse, or unusually large quantity of corpses they now have to deal with.
+
+Format:
+**[Speaker name, Speaker title]:** “[Rumor.]”
+
+FINAL CHECK:
+Before finalizing, make sure:
+
+* There are exactly 5 rumors in each of the 5 sections.
+* The rumors are fun to read aloud.
+* Each section has a distinct worldview and repeating comic pattern.
+* The party is treated unfairly throughout.
+* The rumors are non-canonical and not presented as truth.
+* Each rumor is at least faintly inspired by the session recap.
+* No section includes factual correction, explanation, truth notes, DM-facing notes, or analysis.

prompts/analyze/hostile_rumors/rumors.yml

@@ -0,0 +1,24 @@
+id: dnd.hostile_rumors
+version: "1.0.0"
+default_profile: aion-2
+session_id: "{{ .session_id }}"
+description: D&D session rumor sheet
+inputs:
+  - name: recap
+    required: true
+    content_type: text
+    description: Session recap from the current session
+messages:
+  - role: system
+    content_file: ./rumors.system.md
+  - role: user
+    content_file: ./rumors.user.md
+  - role: user
+    content: |
+      <<<CURRENT_SESSION_RECAP
+      {{input "recap"}}
+      CURRENT_SESSION_RECAP>>>
+output:
+  format: markdown
+  validation_mode: basic
+  repair_attempts: 0

prompts/analyze/scene_map/scene_map.instructions.md

@@ -0,0 +1,77 @@
+INSTRUCTIONS: Your task is to divide the provided transcript into sequential, contiguous, non-overlapping scenes.
+
+A scene is a coherent unit of play. A new scene should usually begin when there is a meaningful change in location, objective, threat, activity, encounter, or mode of play.
+
+Good reasons to start a new scene include:
+- the party moves to a new location
+- a combat encounter begins or ends
+- the combat changes into a substantially different phase
+- the party shifts from combat to exploration, social interaction, planning, travel, or rest
+- a new NPC, faction, threat, or objective becomes central
+- the party completes one immediate goal and begins another
+- a major table-level rules discussion interrupts and materially changes play
+
+Do not start a new scene merely because:
+- the speaker changes
+- a new combat round begins
+- a player asks a brief rules question
+- there is a joke, aside, or short table comment
+- a character takes a routine turn
+- the same encounter continues without a meaningful change in situation
+
+Scene boundary requirements:
+
+1. Cover the transcript from the first relevant segment to the last relevant segment.
+2. Do not leave gaps between scenes.
+3. Do not overlap scenes.
+4. Preserve transcript order.
+5. Use exact segment IDs from the transcript.
+6. Each scene must have a start_segment_id and end_segment_id.
+7. Scene IDs must be sequential: 1, 2, 3, and so on.
+8. Prefer coherent scenes over excessive fragmentation.
+9. Do not merge unrelated scenes just to keep the scene count low.
+10. If a boundary is uncertain, choose the most practical boundary and explain the uncertainty in boundary_note.
+
+For each scene:
+
+- short_title should be brief and factual.
+- primary_mode must use one of the allowed schema values.
+- main_participants should include only the principal characters, NPCs, factions, or groups involved.
+- summary should be factual and compact, usually one to three sentences.
+- boundary_note should explain why the scene begins and ends at those segment IDs.
+- boundary_confidence should be High, Medium, or Low.
+
+Primary mode guidance:
+
+- Use Recap for opening recap, initiative setup, session framing, or immediate continuation from prior events.
+- Use Discussion when the party is primarily discussing options or choosing a course of action.
+- Use Combat when active combat or combat-resolution mechanics dominate.
+- Use Narrative for all other types of non-combat gameplay, including exploration, social interactions, shopping, preparing for battle, downtime, and so on.
+
+Session scope requirements:
+
+- In session_scope.source_description, briefly describe the transcript reviewed.
+- In session_scope.first_segment_id, provide the first segment ID covered by the scene map.
+- In session_scope.last_segment_id, provide the last segment ID covered by the scene map.
+
+Boundary caveats:
+
+In boundary_caveats, list any overall caveats about the scene divisions. Include scenes that could reasonably be split differently, combat phases that were kept together, gradual transitions, or places where map context would have helped.
+
+Important restrictions:
+
+- Do not invent missing events.
+- Do not include comments.
+- Do not include explanatory text outside the JSON object.
+- Do not wrap the JSON in a code block.
+- Return exactly one JSON object.
+
+Before finalizing, check:
+
+- The first scene starts at the first covered transcript segment.
+- The final scene ends at the last covered transcript segment.
+- Scene ranges are sequential and non-overlapping.
+- Every scene has a start_segment_id and end_segment_id.
+- Every summary is factual and compact.
+- Boundary uncertainty is noted rather than hidden.
+- The output is valid JSON and conforms to the schema.

prompts/analyze/scene_map/scene_map.task.md

@@ -0,0 +1,7 @@
+TASK: Generate a structured JSON scene map for the provided D&D session transcript.
+
+This is an intermediate artifact for an automated analysis pipeline. Your task is to divide the transcript into sequential, contiguous, non-overlapping scenes.
+
+Return only valid JSON matching the provided JSON Schema.
+
+Next, you will be provided with detailed instructions with respect to preparation of the strucured JSON scene map.

prompts/analyze/scene_map/scene_map.yml

@@ -0,0 +1,29 @@
+id: dnd.scene_map
+version: "1.0.0"
+default_profile: nemotron-3-ultra
+description: Generates a D&D session scene map, using a transcript as input.
+inputs:
+  - name: transcript
+    required: true
+    content_type: application/json
+    description: Structured recap source transcript
+messages:
+  - role: system
+    content_file: ../common/transcript.system.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ../common/transcript.user.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ./scene_map.task.md
+  - role: user
+    content_file: ./scene_map.instructions.md
+output:
+  format: json
+  validation_mode: json_schema
+  schema_path: pipeline-dnd/analyze/scene_map.schema.json
+  repair_attempts: 2

prompts/analyze/session_recap/recap.instructions.md

@@ -0,0 +1,124 @@
+INSTRUCTIONS: You are an expert assistant helping create permanent campaign logs for a Dungeons & Dragons game.
+
+Your task is to produce polished in-universe session recaps from session transcripts. The recap should be accurate, chronological, useful for archival purposes, and written like a clear campaign chronicle.
+
+Source discipline:
+- The current session transcript is the source of truth for what happened in the session being recapped.
+- Any previous-session recap, campaign notes, glossary, or background material is context only.
+- Do not list something as a development from the current session unless it occurred, was revealed, was decided, or materially changed during the current transcript.
+- Do not invent facts, motivations, names, conditions, locations, relationships, or outcomes that are not supported by the transcript.
+- Do not present player speculation, table theories, or uncertain interpretations as confirmed fact.
+- Use careful language such as “appeared to,” “seemed to,” “the party suspected,” or “it remained unclear” when the transcript does not confirm the point.
+- Do not upgrade uncertain or partial outcomes. For example, do not write “dead,” “unconscious,” “captured,” “possessed,” “defeated,” or “safe” unless the transcript clearly establishes that exact state. Prefer precise observed facts such as “stunned,” “hostile,” “fell from wild shape,” “surrounded,” “apparently compromised,” or “retreated.”
+- If a detail is unclear, inaudible, contradictory, or likely caused by transcription error, either omit it or briefly note the uncertainty rather than guessing.
+
+Before writing the final recap, internally reconstruct the session as a chronological scene ledger. For each major scene, identify:
+- where the scene took place;
+- which PCs, NPCs, factions, monsters, or other important entities were involved;
+- what the party was trying to accomplish;
+- what conflict, decision, discovery, or complication occurred;
+- what changed as a result;
+- what unresolved consequence carried forward.
+
+Do not output this scene ledger. Use it only to make the final recap accurate, complete, and chronological.
+
+Writing requirements:
+- Write entirely in third person.
+  - Correct: “The party entered the ruins.” / “Hrank cast Plant Growth.”
+  - Incorrect: “We entered the ruins.” / “I cast Plant Growth.”
+- Write in past tense unless there is a clear reason not to.
+- Maintain an in-world perspective.
+- Do not refer to “the transcript,” “the players,” “the DM,” “the table,” “the prompt,” or “the recording” in the recap itself.
+- Focus on in-world events, decisions, discoveries, conflicts, and consequences.
+- Exclude out-of-character chatter, jokes, rules lookups, scheduling talk, post-session discussion, and table banter unless they directly changed an in-world decision or outcome.
+- Preserve important proper nouns, locations, factions, magic items, spells, conditions, and NPC names when they are clear from the transcript.
+- Preserve named spells and tactical plans when they materially affected the outcome.
+- When clear from the transcript, explain not only what the party did, but why they chose to do it.
+- For each major action or decision, try to include the consequence: what danger was created or avoided, what resource was spent, what information was learned, what objective was advanced, or what problem remained.
+- Emphasize the most important turning points of the session.
+- If there was a battle, cover the major beats, tactical decisions, reversals, and outcome.
+- If it was a travel or investigation session, cover where the party went, why, what they learned, and what changed.
+
+Output format:
+Use exactly the following five numbered sections.
+
+1. Session Capsule
+
+Write 1 concise paragraph of 4–6 sentences summarizing the overall arc and significance of the session. Include the starting situation, the central conflict or decision, the biggest turning point, and the ending state or cliffhanger.
+
+2. Narrative Summary
+
+Write a chronological recap in polished prose. For a full-length session, write at least 8 paragraphs in the Narrative Summary. Do not use fewer than 8 paragraphs unless the transcript itself is short or uneventful. The Narrative Summary should be detailed enough that someone who missed the session could understand the major decisions, discoveries, combat beats, consequences, and cliffhanger without reading the transcript.
+
+Do not include unnecessary blow-by-blow detail, but do not compress away important complications, failed attempts, major spell uses, tactical plans, reversals, named NPCs, or unresolved dangers.
+
+3. Major Developments
+
+List the most important campaign-level developments from this session as bullet points. Include things like major victories, defeats, revelations, alliances, deaths, discoveries, location changes, recovered artifacts, completed objectives, major tactical outcomes, or newly confirmed threats.
+
+Use 5–10 bullets for a substantial session. Each bullet should state both the development and why it matters.
+
+4. Character Developments
+
+Character Developments should focus on character choices, consequences, relationships, emotional stakes, leadership, failures, risks, vows, losses, changed status, or unresolved personal stakes. Do not use this section as a general action log or resource tracker.
+
+Separately address each player character who participated in the session.
+
+5. Open Threads
+
+Open Threads must reflect the end of the session, not the middle of the session. List unresolved problems, active goals, looming threats, uncertain statuses, and likely next steps that should be remembered for the next session.
+
+Use 5–10 bullets for a substantial session. Be precise about the current state of unresolved tactical situations, endangered characters, hostile NPCs, unfinished combats, unspent resources, and unanswered questions.
+
+Style:
+- Write in English text only. You may use Markdown for headings, lists, bold, and italics. Do not add tables, horizontal 
+- D&D spell names should be italicized.
+- Keep the tone polished, clear, and mildly dramatic, like a campaign chronicle.
+- Prefer concrete events and consequences over vague summary language.
+- Be specific enough to be useful later, but avoid unnecessary blow-by-blow detail.
+- If nothing meaningful belongs in a section, write “None of note.”
+
+For a substantial session, make sure the Narrative Summary covers:
+- the starting situation and immediate aftermath of the previous session;
+- the party’s rest, recovery, and resource constraints;
+- any dangerous prisoners, artifacts, bodies, statues, or unresolved threats being guarded;
+- major NPC revelations or lore discoveries;
+- major defensive, travel, investigation, or tactical plans;
+- the trigger that changed the session’s direction;
+- the main conflict or encounter;
+- the final unresolved state at the end of play.
+
+Mechanical detail:
+- Include game mechanics, spell names, damage, conditions, distances, rolls, and remaining resources only when they materially affect the story, tactical situation, or next-session continuity.
+- Prefer narrative consequence over raw numbers in the Narrative Summary.
+  - Better: “Alan was badly hurt, knocked out of his Quetzalcoatl form, and left stunned near Elric.”
+  - Worse: “Alan took 59 psychic damage and was left at 29 HP,” unless the exact numbers are important for next-session continuity.
+- Raw tactical details such as remaining HP, spell slots, distances, exact enemy counts, or active battlefield effects belong primarily in Open Threads, not Character Developments, unless they reveal something meaningful about the character’s choices or arc.
+
+Named-action attribution:
+- When a named PC or NPC causes a major turning point, identify who did it.
+- Do not write “Grog was hit by a Banishment spell” if the important fact is that Jan used Banishment against Grog.
+
+Rules-state precision:
+- Use D&D condition names carefully. Do not describe a character as unconscious, incapacitated, restrained, grappled, dead, dying, prone, or captured unless that state is clearly established. If the confirmed condition is stunned, say stunned.
+
+Final-state consistency:
+- Before writing Open Threads, silently update the final known status of each major PC, NPC, enemy group, important object, and active location.
+- Do not describe an earlier state as current if a later event changed it.
+- If a character was initially unconscious but later woke, attacked, fled, died, transformed, or was otherwise changed, the Open Threads section must reflect the later state.
+- If the final state is uncertain, describe both the last confirmed fact and the uncertainty.
+- Pay special attention to cliffhanger states at the end of the session.
+
+Out-of-character material:
+- Do not include post-session discussion, player reflections, DM comments, rules debates, or campaign-planning remarks in the main recap.
+- If such material is important for future play, include it only if the requested output format has a separate “Table Notes” or “Meta Notes” section.
+- Otherwise omit it.
+
+Before finalizing, silently check:
+- Did the recap follow the actual chronology of the session?
+- Did it include the major turning points, not just the setup and ending?
+- Did it preserve precise tactical states such as stunned, unconscious, dead, captured, transformed, restrained, or fleeing?
+- Did it avoid presenting speculation as fact?
+- Did it include the major consequences and open threads?
+- Did each meaningful PC spotlight receive attention?
+- Did it avoid out-of-character table discussion?

prompts/analyze/session_recap/recap.task.md

@@ -0,0 +1,5 @@
+TASK: Produce a polished in-universe recap of the provided Dungeons & Dragons session transcript.
+
+A summary of the previous D&D session may also be provided for context. If present, use it to understand the party’s starting situation, ongoing plot threads, recurring NPCs, and target recap style. Do not treat the previous recap as evidence of what happened in the current session.
+
+Next, you will be provided with detailed instructions with respect to preparation of the polished in-universe recap.

prompts/analyze/session_recap/recap.yml

@@ -0,0 +1,38 @@
+id: dnd.session_recap
+version: "2.0.0"
+default_profile: claude-fable-latest
+session_id: "{{ .session_id }}"
+description: D&D session recap prompt.
+inputs:
+  - name: transcript
+    required: true
+    content_type: application/json
+    description: Structured recap source transcript
+  - name: recap
+    required: false
+    content_type: text
+    description: Optional previous session recap
+messages:
+  - role: system
+    content_file: ../common/transcript.system.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ../common/transcript.user.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ../common/transcript.previous_recap.md
+    cache_control:
+      type: ephemeral
+      ttl: 1h
+  - role: user
+    content_file: ./recap.task.md
+  - role: user
+    content_file: ./recap.instructions.md
+output:
+  format: markdown
+  validation_mode: basic
+  repair_attempts: 0

prompts/trim/bounds.system.md

@@ -0,0 +1,64 @@
+You are a transcript boundary detector for a Dungeons & Dragons actual-play session.
+
+Your task is to identify the contiguous range of transcript segments that covers the actual gameplay session, excluding pre-session setup chatter and post-session banter.
+
+The transcript consists of speaker segments. Each segment has an `id`, a `speaker`, and text. Segment IDs are stable and should be used as the boundary markers.
+
+Definitions:
+
+Actual gameplay includes:
+- the DM beginning the session recap or scene setup;
+- the DM narrating the current in-world situation;
+- players discussing in-character decisions, tactics, exploration, combat, travel, spellcasting, roleplay, or plans;
+- rules discussion that occurs during gameplay and affects in-world decisions;
+- end-of-session narration, cliffhanger setup, combat state, or next-session handoff.
+
+Pre-session chatter includes:
+- greetings, microphone checks, recording setup, waiting for players, food/drink talk, scheduling before play begins, casual conversation, jokes, or unrelated discussion before the DM begins the session.
+
+Post-session banter includes:
+- discussion after the DM has clearly ended play;
+- scheduling the next session;
+- post-session rules debate that does not change the in-world outcome;
+- player commentary, jokes, debrief, technical discussion, or unrelated conversation after the gameplay endpoint.
+
+Boundary rules:
+- Identify one continuous inclusive range: `start_segment_id` through `end_segment_id`.
+- The `start_segment_id` should be the first segment that is meaningfully part of the session: usually the DM’s opening recap, scene-setting, or the first in-world/player decision after pre-session chatter.
+- The `end_segment_id` should be the last segment that is meaningfully part of gameplay: usually the final in-world narration, final adjudicated action, combat state, cliffhanger, or “we’ll pick up there next time” moment.
+- If a segment contains both chatter and gameplay, include it if any meaningful gameplay occurs in that segment.
+- Do not exclude mid-session breaks, jokes, or tangents if they occur between the start and end of gameplay. The goal is to remove only the pre-session prefix and post-session suffix.
+- Do not skip the opening recap, planning, recovery, shopping, travel preparation, or rules adjudication merely because it is less dramatic than later combat. If it occurs after gameplay has begun and affects the session’s state, it belongs inside the range.
+- When uncertain, choose a slightly wider range. It is better to include a small amount of chatter than to exclude actual gameplay.
+- Players may provide spoken indications that the gameplay is beginning or ending. For example, speech such as "Are we ready to get started?" is a strong indication that gameplay is beginning. Likewise, speech such as "Should we end it here?" is a strong indication that gameplay is ending.
+- Do not infer missing segment IDs. Only use IDs that appear in the transcript.
+- Do not summarize the session.
+
+Return only valid JSON. Do not include markdown, commentary, or prose outside the JSON object.
+
+Use this JSON shape:
+
+{
+  "start_segment_id": 0,
+  "end_segment_id": 0,
+  "confidence": "high",
+  "start_boundary_reason": "",
+  "end_boundary_reason": "",
+  "start_evidence": {
+    "segment_id": 0,
+    "speaker": "",
+    "text_excerpt": ""
+  },
+  "end_evidence": {
+    "segment_id": 0,
+    "speaker": "",
+    "text_excerpt": ""
+  },
+  "excluded_prefix_summary": "",
+  "excluded_suffix_summary": "",
+  "warnings": []
+}
+
+The `confidence` value must be one of: "high", "medium", or "low".
+
+Keep evidence excerpts short. Quote only enough text to justify the boundary.

prompts/trim/bounds.user.md

@@ -0,0 +1,13 @@
+Identify the inclusive segment ID range that covers the actual gameplay session in the transcript below.
+
+Use the first gameplay segment as `start_segment_id`.
+Use the final gameplay segment as `end_segment_id`.
+
+Remember:
+- Include the DM’s opening recap or scene-setting if present.
+- Include the final in-world narration, unresolved combat state, cliffhanger, or explicit handoff to next session.
+- Exclude only pre-session chatter before gameplay begins and post-session banter after gameplay ends.
+- Return only valid JSON using the required schema.
+- Do not skip the opening recap, planning, recovery, shopping, travel preparation, or rules adjudication merely because it is less dramatic than later combat. If it occurs after gameplay has begun and affects the session’s state, it belongs inside the range.
+
+<<< TRANSCRIPT TO FOLLOW >>>

prompts/trim/bounds.yml

@@ -0,0 +1,23 @@
+id: dnd.session_bounds
+version: "2.0.0"
+default_profile: gemini-flash-latest
+session_id: "{{ .session_id }}"
+description: Review D&D session transcript and identify starting and ending segments for actual gameplay.
+inputs:
+  - name: transcript
+    required: true
+    content_type: application/json
+    description: Structured recap source transcript
+messages:
+  - role: system
+    content_file: ./bounds.system.md
+  - role: user
+    content_file: ./bounds.user.md
+  - role: user
+    content: |
+      {{ input "transcript" }}
+output:
+  format: json
+  validation_mode: json_schema
+  schema_path: automation-pipeline-dnd/trim/bounds.schema.json
+  repair_attempts: 0

schemas/analyze/scene_map.schema.json

@@ -0,0 +1,122 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "title": "D&D Session Scene Map",
+  "description": "Structured scene segmentation for a D&D session transcript.",
+  "type": "object",
+  "additionalProperties": false,
+  "required": [
+    "artifact_type",
+    "session_scope",
+    "scenes",
+    "boundary_caveats"
+  ],
+  "properties": {
+    "artifact_type": {
+      "type": "string",
+      "const": "current_session_scene_map"
+    },
+    "session_scope": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": [
+        "source_description",
+        "first_segment_id",
+        "last_segment_id"
+      ],
+      "properties": {
+        "source_description": {
+          "type": "string",
+          "description": "Brief description of the transcript or source material reviewed."
+        },
+        "first_segment_id": {
+          "type": "integer",
+          "description": "The first transcript segment covered by the scene map."
+        },
+        "last_segment_id": {
+          "type": "integer",
+          "description": "The final transcript segment covered by the scene map."
+        }
+      }
+    },
+    "scenes": {
+      "type": "array",
+      "description": "Sequential, contiguous, non-overlapping scene divisions covering the transcript.",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": [
+          "scene_id",
+          "start_segment_id",
+          "end_segment_id",
+          "short_title",
+          "primary_mode",
+          "main_participants",
+          "summary",
+          "boundary_note",
+          "boundary_confidence"
+        ],
+        "properties": {
+          "scene_id": {
+            "type": "string",
+            "description": "Stable sequential scene identifier.",
+            "pattern": "[0-9]$"
+          },
+          "start_segment_id": {
+            "type": "integer",
+            "description": "Exact segment ID where this scene begins."
+          },
+          "end_segment_id": {
+            "type": "integer",
+            "description": "Exact segment ID where this scene ends."
+          },
+          "short_title": {
+            "type": "string",
+            "description": "Compact descriptive title for the scene."
+          },
+          "primary_mode": {
+            "type": "string",
+            "enum": [
+              "Recap",
+              "Planning",
+              "Combat",
+              "Narrative"
+            ],
+            "description": "The dominant mode of play in the scene."
+          },
+          "main_participants": {
+            "type": "array",
+            "description": "Principal player characters, NPCs, factions, or groups involved in the scene.",
+            "items": {
+              "type": "string"
+            }
+          },
+          "summary": {
+            "type": "string",
+            "description": "One- to three-sentence factual summary of what happened in the scene."
+          },
+          "boundary_note": {
+            "type": "string",
+            "description": "Brief explanation of why the scene starts and ends at these segment boundaries."
+          },
+          "boundary_confidence": {
+            "type": "string",
+            "enum": [
+              "High",
+              "Medium",
+              "Low"
+            ],
+            "description": "Confidence that the chosen scene boundaries are natural and useful."
+          }
+        }
+      }
+    },
+    "boundary_caveats": {
+      "type": "array",
+      "description": "Overall caveats about scene segmentation, ambiguous transitions, map-dependent context, or alternative reasonable divisions.",
+      "items": {
+        "type": "string"
+      }
+    }
+  }
+}

schemas/trim/bounds.schema.json

@@ -0,0 +1,122 @@
+{
+  "schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "description": "Identifies the inclusive range of transcript segment IDs that covers the actual Dungeons & Dragons gameplay session, excluding pre-session chatter and post-session banter.",
+  "required": [
+    "trim_action",
+    "start_segment_id",
+    "end_segment_id",
+    "confidence",
+    "start_boundary_reason",
+    "end_boundary_reason",
+    "start_evidence",
+    "end_evidence",
+    "excluded_prefix_summary",
+    "excluded_suffix_summary",
+    "warnings"
+  ],
+  "additionalProperties": false,
+  "properties": {
+    "trim_action": {
+      "type": "string",
+      "description": "Whether the transcript should be trimmed, appears to need no trimming, or is too ambiguous for confident trimming.",
+      "enum": [
+        "trim",
+        "no_trim_needed",
+        "uncertain"
+      ]
+    },
+    "start_segment_id": {
+      "type": "integer",
+      "description": "The inclusive segment ID where actual gameplay begins. This should be the first meaningful gameplay segment, such as the DM's opening recap, scene-setting, or the first in-world decision.",
+      "minimum": 0
+    },
+    "end_segment_id": {
+      "type": "integer",
+      "description": "The inclusive segment ID where actual gameplay ends. This should be the final meaningful gameplay segment, such as final in-world narration, an adjudicated action, a cliffhanger, or a next-session handoff.",
+      "minimum": 0
+    },
+    "confidence": {
+      "type": "string",
+      "description": "Confidence in the selected boundaries.",
+      "enum": [
+        "high",
+        "medium",
+        "low"
+      ]
+    },
+    "start_boundary_reason": {
+      "type": "string",
+      "description": "Brief explanation of why the selected start segment is the first gameplay segment."
+    },
+    "end_boundary_reason": {
+      "type": "string",
+      "description": "Brief explanation of why the selected end segment is the last gameplay segment."
+    },
+    "start_evidence": {
+      "type": "object",
+      "description": "Short evidence supporting the selected start boundary.",
+      "properties": {
+        "segment_id": {
+          "type": "integer",
+          "description": "The segment ID used as evidence for the start boundary.",
+          "minimum": 0
+        },
+        "speaker": {
+          "type": "string",
+          "description": "The speaker of the evidence segment."
+        },
+        "text_excerpt": {
+          "type": "string",
+          "description": "A short excerpt from the evidence segment showing why gameplay begins here."
+        }
+      },
+      "required": [
+        "segment_id",
+        "speaker",
+        "text_excerpt"
+      ],
+      "additionalProperties": false
+    },
+    "end_evidence": {
+      "type": "object",
+      "description": "Short evidence supporting the selected end boundary.",
+      "properties": {
+        "segment_id": {
+          "type": "integer",
+          "description": "The segment ID used as evidence for the end boundary.",
+          "minimum": 0
+        },
+        "speaker": {
+          "type": "string",
+          "description": "The speaker of the evidence segment."
+        },
+        "text_excerpt": {
+          "type": "string",
+          "description": "A short excerpt from the evidence segment showing why gameplay ends here."
+        }
+      },
+      "required": [
+        "segment_id",
+        "speaker",
+        "text_excerpt"
+      ],
+      "additionalProperties": false
+    },
+    "excluded_prefix_summary": {
+      "type": "string",
+      "description": "Brief summary of the material excluded before the start segment. Use an empty string if no prefix was excluded."
+    },
+    "excluded_suffix_summary": {
+      "type": "string",
+      "description": "Brief summary of the material excluded after the end segment. Use an empty string if no suffix was excluded."
+    },
+    "warnings": {
+      "type": "array",
+      "description": "Any caveats, ambiguities, or reasons the boundary selection may need human review.",
+      "items": {
+        "type": "string"
+      }
+    }
+  }
+}