Interface Learnings

Engine: PMLC = python3 <skills-root>/learnings-coach/scripts/pmlc.py (this skill needs the learnings-coach skill installed alongside it). Needs clusters.json.

Gather

PMLC interfaces <ws> --top 25

Each pair has shared findings (both entities named on the same finding) and cross-entity causal mentions (a finding owned by one entity names a factor owned by the other in its causal explanation). Rank interfaces by both. Take the top eight to twelve.

For each interface read the examples: PMLC select <ws> --entity <A> <B> --entity-mode all --brief (the findings both entities share), then the full text of the highest-rated ones.

For each interface, write

• The seam: what passes between the two functions (a work order, a part, a test, an alarm, a decision, a document).
• What went wrong at the seam: assumption, gap in ownership, missing hand-back, conflicting priorities, authority that sat with neither.
• Why neither side saw it on its own.
• The learning for each side in one sentence.
• A joint fix: a hold point, a shared check, a single owner, a handover record, a joint review. Name the control tier.
• A leading indicator both sides can see.
• Evidence trail.

Output

Write the pack <run>/outputs/interface-learnings.json (schema: learnings-coach/references/pack-schema.md), then:

PMLC check-pack <ws> <run>/outputs/interface-learnings.json   # fix every problem it lists
PMLC pack <ws> <run>/outputs/interface-learnings.json         # -> interface-learnings.html (primary) and .md

Sections and the kinds that fill them:

• Matrix: a table with columns Interface, Shared findings and Cross-entity causal mentions. Use the top pairs from interfaces, with their examples as row_cites.
• One section per interface, with id = the pair (e.g. f02-f06):
  • para items with group = "The seam", "What went wrong at the seam" or "Why neither side saw it".
  • A learning for each side, with owner = that side.
  • The joint fix as a recommendation, with a hoc_tier and an indicator both sides can see.
• Who must meet whom: a closing section of check items.

Put data-quality corrections in data_quality, findings left out on purpose in exclusions, and same-condition findings in supporting_cites.

{"kind": "learning", "group": "The learning for each side", "lesson": "Stores: …", "owner": "Materials Management (F06)",
 "cites": ["FND-…"]}