investigator-coach
Skill0 starsChanged 29 days agoMITView on GitHub
Coach incident investigators and HSE professionals using a ProbeMatrix investigation export - evidence gaps and conflicts, key questions and suggested interviews still to pursue, causal hypotheses to test, and method lessons on reaching supervisory and organisational causes. Fills the export's Evidence to Collect, Key Questions, Suggested Interviews and Causal Hypothesis columns when they are empty. Use when someone asks how to deepen or strengthen an investigation, what evidence is missing, who else to interview, how good the investigation is, or wants investigator training material from a real case.
In this repository24 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- auditor-assurance-packSkill
- barrier-learningsSkill
- board-briefSkill
- causal-mapSkill
- cluster-learning-dossierSkill
- emergency-response-learningsSkill
- engineering-integrity-learningsSkill
- frontline-learning-cardsSkill
- interface-learningsSkill
- learning-clustersSkill
- learnings-coachSkill
- learnings-registerSkill
- manager-coachSkill
- organisation-learning-mapSkill
- precursor-timelineSkill
- report-comparisonSkill
- safety-committee-guideSkill
- site-leadership-coachSkill
- supervisor-coachSkill
- supply-chain-contractor-learningsSkill
- top-management-briefSkill
- training-designerSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Read the incident narrative in <ws>/incident.json and the evidence descriptions in <ws>/evidence.json.
- Very specific: line 31
Write the pack <run>/outputs/investigator-coach.json (schema: learnings-coach/references/pack-schema.md), then:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Investigator Coach
Audience: investigators, HSE advisers and process-safety engineers who run or review investigations.
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 and export-schema.md.
Gather
PMLC profile <ws> # shows whether follow-up columns are empty
PMLC select <ws> --top-rated --format json --out <ws>/top.jsonRead the incident narrative in <ws>/incident.json and the evidence descriptions in <ws>/evidence.json.
Write
- Evidence base audit: evidence items by type; which kinds of evidence are absent (e.g. no maintenance-fitter statement, no procurement records, no calibration master data); conflicts between sources (dates, times, identifiers) and how they were resolved or not.
- Follow-up per high-rated finding (one table row each): evidence to collect · key questions · suggested interviews (by role) · causal hypothesis to test. Keep it specific to the finding's evidence.
- Interview plan (follow "Interviews in a just culture" in
learnings-coach/references/learning-principles.md: non-disciplinary, a representative present, local rationality, no leading questions): roles to interview that the evidence never heard from, ordered by how much they could change the analysis, each with five questions. - Method lessons: where the analysis reached L3–L4 and how; where findings may be coded at too low a level (an organisational condition recorded as an individual act); where a finding rests on a single source.
- Training use: three exercises from this case for investigator training (e.g. "move this L1 finding to the level where it belongs").
Output
Write the pack <run>/outputs/investigator-coach.json (schema: learnings-coach/references/pack-schema.md), then:
PMLC check-pack <ws> <run>/outputs/investigator-coach.json
PMLC pack <ws> <run>/outputs/investigator-coach.json # -> investigator-coach.html, .md and follow-up.jsonSections → kinds:
| Section | Kind |
| 1 Evidence audit | table (evidence type, items) + para |
| 2 Follow-up | the table with "id": "follow-up" and columns Finding, Evidence to collect, Key questions, Suggested interviews, Causal hypothesis, one row per condition, with that condition's high-rated Finding IDs (comma-separated) in the first cell and row_cites |
| 3 Interview plan | question, to the role |
| 4 Method lessons | para |
| 5 Training use | scenario |
pack writes follow-up.json from the follow-up table (one entry per finding), ready to paste back into the investigation. Use data_quality for export problems and exclusions for findings left out on purpose. When one finding in a shared row needs its own note, add finding_notes ({"FND-…": "note"}) to the table.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

