Engineering & Integrity Learnings

Audience: the Learning Clusters tagged engineering (in the default taxonomy F04 instrumentation, control and calibration, F05 asset integrity, corrosion and inspection, F07 engineering, design, documents and MOC, and F03 electrical, for hazardous-area equipment).

Engine: PMLC = python3 <skills-root>/learnings-coach/scripts/pmlc.py (this skill needs the learnings-coach skill installed alongside it). Read learnings-coach/references/learning-principles.md.

Gather

PMLC select <ws> --tag engineering --brief
PMLC select <ws> --tag engineering --level 3 4 --top-rated
PMLC select <ws> --text "specification|drawing|P&ID|MOC|change|inspection|thickness|corrosion|calibrat|interlock|test|design" --brief
PMLC timeline <ws> --before incident

Write

1. Technical learnings by theme: specification and substitution of parts; testing after maintenance; change management; inspection and corrosion programmes; documents and as-built records; detection, alarm and shutdown design; hazardous-area equipment integrity; ageing equipment. Only themes the evidence supports.
2. For each learning: the technical condition, how it persisted (L3–L4), the barrier it weakened, a recommendation with HoC tier (engineering controls first where credible), and a technical leading indicator.
3. Engineering decisions pending: questions the technical authority must settle (e.g. test pressure policy after pressure-boundary work, acceptance of substitute components), stated as decisions, not answers.
4. Data and records: what technical records were missing or unreliable, and what the minimum record should be.

Output

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

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

Sections → kinds:

1. For each learning, write three items with the same group (the theme):
   • a learning: the technical condition and how it persisted;
   • a barrier for the barrier it weakened (type, owner, state, failure_mode, hoc_tier, fix, indicator);
   • a recommendation, engineering controls first where credible.
2. decision items (options, no answers), with owner = the technical authority.
3. check items stating the minimum record, with when = the point in the work.

Set "scope": {"tags": ["engineering"]}. Put data-quality corrections in data_quality, findings left out on purpose in exclusions, and same-condition findings in supporting_cites.

{"kind": "barrier", "group": "Specification and substitution", "barrier": "…", "type": "prevention", "owner": "…",
 "state": "absent", "failure_mode": "…", "hoc_tier": "engineering", "fix": "…", "indicator": "…", "cites": ["FND-…"]}

Numbers rule. Do not specify engineering values (pressures, thicknesses, intervals, percentages, durations) beyond what the cited records state; never add a figure from outside the export; make them decision items for the technical authority.