Document chunk plan caching and provenance

docs/policy/architecture.md

@@ -105,6 +105,14 @@ Stage ownership is explicit:
 - normalize modules reconcile merged output;
 - output modules encode accepted results and run outcomes into logical files.
 
+Chunk modules produce source-addressed chunk plans rather than materialized
+chunks. The framework validates and materializes those plans into the generic
+chunk representation before chunk validation and lane execution. Plan reuse is
+therefore independent of the configured pipeline, module options, references,
+lanes, validators, and LLM profile: the canonical source digest selects the
+plan, while the current run still applies its configured chunk validators to
+the materialized chunks.
+
 The framework owns orchestration and handoff provenance. Modules return logical
 results and warnings; they do not own CLI reporting, workspace paths, durable
 file placement, checkpoints, or diagnostics.
@@ -191,6 +199,12 @@ surfaces with separate ownership:
 - debug artifacts are opt-in inspection data and may contain sensitive source,
   prompt, reference, and model-output content.
 
+Chunk-plan cache state is an additional independent surface. It is keyed only
+by canonical source digest, is not rooted under `workspace.directory`, and is
+not a checkpoint or a diagnostic. A cache record is atomically replaced as one
+complete plan envelope; it has no history, locking, or rollback interface.
+Invalid records are recoverable cache misses rather than pipeline state.
+
 Writes of durable state are atomic where practical. Paths for writes, moves,
 overwrites, and deletion must be narrow and explicit. Cleanup that can lose data
 is opt-in.