graph-lint
Skill2 starsChanged 49 days agoMITView on GitHub
Health check for the graph: dead links, incomplete Signals, orphans, stale data — proposes fixes, never rewires links without approval. Use on /graph-lint or after large edit batches/imports.
In this repository31 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- affinitySkill
- ai-personaSkill
- analytics-syncSkill
- backlogSkill
- catalogSkill
- cold-startSkill
- contradictionsSkill
- demo-dataSkill
- dovetail-syncSkill
- exportSkill
- extract-findingsSkill
- feature-panelSkill
- forget-participantSkill
- interview-guideSkill
- journey-mapSkill
- opportunity-treeSkill
- persona-avatarSkill
- persona-querySkill
- persona-talkSkill
- persona-voiceSkill
- persona-workshopSkill
- prdSkill
- prioritizeSkill
- researcherSkill
- retentionSkill
- source-syncSkill
- subject-accessSkill
- welcomeSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Run the mechanical linter, interpret its findings, and propose fixes.
- Very specific: line 21
Run python3 scripts/graph_lint.py and show the output grouped as the script prints it (ERRORS / WARNINGS / INFO).
- Very specific: line 23
If --fix (or the user asks): propose fixes per the approval rules below.
- Distinctive: line 32
If the user declines a fix, leave it and don't re-propose it every run — note it as "known, accepted by researcher" in the summary.
- Distinctive: line 36
Never edits files during the report phase.
- Distinctive: line 37
Never deletes a Signal/Evidence file (even an orphan) — worst case it proposes archiving, user decides.
- Very specific: line 38
Never touches Transcripts/ content.
≈ “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
graph-lint — Graph health check
Run the mechanical linter, interpret its findings, and propose fixes. This skill reports and proposes — the researcher owns the graph.
Trigger
/graph-lint ← full report
/graph-lint --fix ← report, then walk through proposed fixes (with approval)Also worth suggesting (not auto-running) after a big import (/dovetail-sync, /source-sync, /extract-findings --all).
Workflow
- Run
python3 scripts/graph_lint.pyand show the output grouped as the script prints it (ERRORS / WARNINGS / INFO). - Interpret briefly — e.g. orphan Signals may simply belong to a persona not yet written (offer
/persona-workshop), stale INFO items are candidates for a refresh round (/researcherfor Evidence, real interviews for Signals). - If
--fix(or the user asks): propose fixes per the approval rules below.
Approval rules — the researcher stays in control
Hand-written links are the researcher's material. The AI never silently rewires the graph:
- Trivial mechanical fixes — a link whose intended target is obvious but the path is broken (missing URL-encoding, a typo'd filename, a file that was renamed and every other reference already points to the new name): batch them into ONE list, show old → new for each, apply only after a single explicit yes.
- Meaning-changing fixes — adding a link that didn't exist (e.g. wiring an orphan Signal to a Persona), removing a link, changing which entity something points to, or editing
evidences:lists: propose each one individually with a one-line justification citing the content that supports it. These change what the graph claims, so they are individually approved, never batched. - Never "fix" a contradiction — conflicting Signals are data. If lint output or content review surfaces a semantic conflict, point to
/contradictionsinstead of editing anything. - If the user declines a fix, leave it and don't re-propose it every run — note it as "known, accepted by researcher" in the summary.
What this skill never does
- Never edits files during the report phase.
- Never deletes a Signal/Evidence file (even an orphan) — worst case it proposes archiving, user decides.
- Never touches
Transcripts/content.
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.

