diff --git a/assets/dnd/shared/prompts/common-dnd-extraction-evidence.md b/assets/dnd/shared/prompts/common-dnd-extraction-evidence.md
index 439bce3..ad570a9 100644
--- a/assets/dnd/shared/prompts/common-dnd-extraction-evidence.md
+++ b/assets/dnd/shared/prompts/common-dnd-extraction-evidence.md
@@ -1,7 +1,6 @@
 Transcript units are the only evidence for extracted events and factual claims.
 Every reported factual claim must be supported by cited transcript units. Use
-integer `start_unit_id` and `end_unit_id` values from the transcript. Omit
-`source_id`; Notarius assigns the current source identity.
+integer `start_unit_id` and `end_unit_id` values from the transcript.
 
 When supporting evidence is non-contiguous, use multiple narrow ranges rather
 than a broad range that bridges unrelated conversation.
diff --git a/assets/dnd/shared/prompts/common-dnd-system.md b/assets/dnd/shared/prompts/common-dnd-system.md
index 4f7cfbe..977680e 100644
--- a/assets/dnd/shared/prompts/common-dnd-system.md
+++ b/assets/dnd/shared/prompts/common-dnd-system.md
@@ -1,7 +1,5 @@
 You process Dungeons & Dragons gameplay transcripts.
 
-Rely only on the supplied inputs. They may contain transcription errors,
-repeated lines, incomplete sentences, and misheard proper nouns.
+As input, you will receive one or more portions of a transcript. The transcript may contain transcription errors, repeated lines, incomplete sentences, and misheard proper nouns.
 
-Return exactly one JSON object that conforms to the configured response schema,
-with no explanatory prose.
+Return exactly one JSON object that conforms to the configured response schema, with no explanatory prose.
diff --git a/assets/dnd/shared/prompts/common-dnd-transcript-chunk.md b/assets/dnd/shared/prompts/common-dnd-transcript-chunk.md
index ef77a71..e8ce336 100644
--- a/assets/dnd/shared/prompts/common-dnd-transcript-chunk.md
+++ b/assets/dnd/shared/prompts/common-dnd-transcript-chunk.md
@@ -1,5 +1,3 @@
-One extraction chunk from a Dungeons & Dragons gameplay transcript is provided
-below. Report and infer only what is within this chunk. Its unit IDs retain
-their source-wide meaning.
+One extraction chunk from a Dungeons & Dragons gameplay transcript is provided below. Report and infer only what is within this chunk. Its unit IDs retain their source-wide meaning.
 
 {{ input "transcript" }}
diff --git a/assets/dnd/shared/prompts/common-dnd-transcript-full.md b/assets/dnd/shared/prompts/common-dnd-transcript-full.md
index 84744f4..7ff0ff9 100644
--- a/assets/dnd/shared/prompts/common-dnd-transcript-full.md
+++ b/assets/dnd/shared/prompts/common-dnd-transcript-full.md
@@ -1,4 +1,3 @@
-The complete ordered transcript of this Dungeons & Dragons gameplay session is
-provided below. It may contain multiple scenes.
+The complete ordered transcript of this Dungeons & Dragons gameplay session is provided below.
 
 {{ input "transcript" }}
diff --git a/assets/generic/normalize/deduplication/candidates.md b/assets/generic/normalize/deduplication/candidates.md
new file mode 100644
index 0000000..d08b3f1
--- /dev/null
+++ b/assets/generic/normalize/deduplication/candidates.md
@@ -0,0 +1,3 @@
+Candidates for identity comparison:
+
+{{ input "candidates" }}
diff --git a/assets/generic/normalize/deduplication/instructions.md b/assets/generic/normalize/deduplication/instructions.md
new file mode 100644
index 0000000..6934c9e
--- /dev/null
+++ b/assets/generic/normalize/deduplication/instructions.md
@@ -0,0 +1,5 @@
+Your task is to identify duplicate entities withi the provided list of candidates.
+
+Please review the listed candidates and their underlying evidence, and determine whether any of the listed candidates refer to the same underlying entity. Preserve distinct candidates that appear to refer to different underlying entities, even when their names are similar.
+
+When selecting a canonical display name, prefer a complete, stable proper name over an abbreviation. Prefer an unadorned proper name over that name plus additional descriptors, unless the underlying evidence stablishes the descriptors as part of the entity's name. A longer display name is not inherently more canonical.