path-scope-example
Skill0 starsChanged 6 months agoMITView on GitHub
The kit's reference example of a PATH-SCOPED skill (ADR-0010) — copy its `paths:` frontmatter as the template for your own directory-scoped skills. Its `paths:` block, not this description, is what loads it: inside the decisions folder it auto-loads as a light ADR-authoring reminder (next free ADR-NNNN, the template to start from, the same-response linking rule) and everywhere else it stays silent.
In this repository32 other files it gives its agents.
- _00-Project-Management/CLAUDE.mdCLAUDE.md
- autopilotSkill
- close-out-storySkill
- close-phaseSkill
- coreSkill
- critiqueSkill
- curate-toolkitSkill
- documentSkill
- draft-epicSkill
- draft-okrsSkill
- draft-prdSkill
- execute-batch-parallelSkill
- execute-batchSkill
- execute-storySkill
- execution-strategistSkill
- fill-claude-mdSkill
- installSkill
- modeSkill
- monthly-retroSkill
- peer-reviewSkill
- plan-driftSkill
- refine-backlogSkill
- reflectSkill
- run-testplanSkill
- session-startSkill
- split-into-featuresSkill
- split-into-storiesSkill
- start-phaseSkill
- triage-backlogSkill
- updateSkill
- weekly-monitorSkill
- write-outcomesSkill
The file
Files
path-scope-example (reference: directory-scoped skill)
This skill is the kit's worked example of paths: scoping (ADR-0010, proven by STORY-17.2.01). Unlike every other Tandem skill — which is globally relevant and explicitly slash-invoked — this one is bound by its paths: frontmatter to _00-Project-Management/40-Decisions/, so Claude auto-loads it only while reading/writing files in that directory. Outside the decisions folder it stays silent.
It is intentionally minimal and doubles as a genuine, directory-local helper: a quick ADR-authoring checklist.
When this activates
Auto-loads when you touch a file under _00-Project-Management/40-Decisions/ (e.g. creating ADR-0077-*.md). It does not auto-load when you're working in 32-Stories/, skills/, or anywhere else — that's the whole point of path-scoping.
ADR-authoring reminder (the local concern)
- Number sequentially across the whole project. Glob
ADR-*.mdin this folder, take the maxNNNN, add 1. No folder grouping. - Start from the template —
91-Templates/ADR.template.md. Don't redraft headings from memory. - Record the decision in the same response as the work that forced it (ADR-on-the-spot).
- Link back — add the new
ADR-NNNNto the originating story'sdecisions:frontmatter array. - Commit, don't hedge — an ADR commits to one option; list the rejected ones under
## Alternatives considered.
How to reuse this pattern
To make one of your own skills directory-scoped, copy the paths: block above and point it at your directory's glob (repo-relative, ** recursion, brace expansion — same format as .claude/rules/ path-specific rules). Remember paths: is additive with description: — it constrains where a description-match may fire; it is not a security control. See 90-Standards/CLAUDE-CODE-CONFIG.md §2.3.1 for when to path-scope vs description-match.
Portability note: paths: is a Claude Code loader convention, not part of the minimal skill-frontmatter schema (name + description). Hosts and validators that don't recognise it ignore unknown frontmatter keys, so on such hosts (e.g. Claude Cowork surfaces that only description-match) this skill degrades gracefully to an ordinary description-matched skill — nothing breaks, it just loses the auto-load-on-directory behaviour.
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.

