papyrus-update-metadata

name	papyrus-update-metadata
description	Atomic — patch mutable fields of an existing memory

papyrus-update-metadata

Patch one memory's mutable fields (body, title, tags, confidence, scope, status, source, links). Immutable fields (id, type, created_at) rejected.

Atomicity

• (a) Indivisible: single update_need call, filelocked
• (b) Input: id + patch dict. Output: updated Need or KeyError/ValueError
• (c) Objective reward: reload by id, check patched fields match input
• (d) Reusable: invoked whenever metadata needs change (confidence elevation, review_after refresh, tag reorganisation)
• (e) No interference: operates on one memory under filelock

Input / Output

input:
  id:     non-empty string
  patch:  dict with any subset of {body, title, tags, confidence, scope, status, source, links}
output:
  side effect: rewrites memory/<plural>.rst with patched need; updated_at refreshed
  stdout: "Updated <id>"
  exit: 0 on success; non-zero on unknown id or immutable-field attempt

Success criterion

papyrus get <id>   # exits 0 and shows patched fields

Workflow

MCP tool memory_update:

{"id": "FACT_x", "body": "corrected", "confidence": "high"}

Composition examples

Elevate confidence after verification:

1. papyrus-query + papyrus-drill — confirm memory holds
2. Verify independently (code inspection, test run)
3. papyrus-update-metadata set confidence=high
4. Optionally: papyrus-promote to higher scope

Extend review window:

1. papyrus-detect-stale returns overdue fact X
2. Verify X still true
3. papyrus-update-metadata reset review_after to now + 180 days

See also

• papyrus-deprecate — for retirement (distinct operation)
• shared/apply-memory.md — when conflict justifies update vs supersede