Repository navigation
Enforce the first-party skill authoring rules in CI #678
Description
🤖 Generated by the Agentic Engineer (interactive Claude Code session)
Problem
#641 asked for the authoring rules for first-party skills to be enforced, not only written down. #677 delivered the rules (the skill-lifecycle skill's authoring.md) and CI already validates each skill folder against the open skill format, the licence records and the provenance of external skills. Four rules have no check, so they hold only when an author remembers them:
- the entry file stays under 500 lines;
- reference files are linked from the entry file and sit one level deep;
- the description stays within a stated length unless the change says why (the coverage audit found 30 of 94 listed skills shown by name only, Spike: evaluate whether agents select and apply the right skills #642);
- a first-party skill's acceptance scenarios exist somewhere a suite can read them.
Proposed direction
Extend the existing validation script for first-party skills only (external skills are upstream-owned and are not edited here):
- fail on an entry file over 500 lines and on a reference file nested deeper than one level or not linked from the entry file;
- report each first-party description's length, and fail above a threshold chosen from the current distribution, with an explicit allow-list for the ones that have a recorded reason;
- decide with Measure skill activation and adherence with a claude plugin eval suite #651 where acceptance scenarios live, and check that each first-party skill has at least two.
Start by measuring the current first-party skills against 1 and 2, so the thresholds are set from data and no existing skill fails on day one without a decision.
Acceptance criteria
- CI fails a first-party skill whose entry file exceeds 500 lines or whose references are nested or unlinked, with a self-test for each case.
- Description lengths of first-party skills are reported in CI, and the threshold and its exceptions are recorded.
- The scenario check is added, or this issue records that Measure skill activation and adherence with a claude plugin eval suite #651 owns it.
-
docs/skill-lifecycle.mdlists the new checks under what CI checks.
Size
Small to medium.
Follows from #641 and #677. Related: #651 (activation suite), #642 (listing budget).
🤖 Generated with Claude Code
Activity
coderabbitai commentedon Nov 10, 2026
🔗 Related PRs
#619 - feat(skills): own maintained sources in their plugins [merged]
📝 Issue Planner
Check the box below or use the
@coderabbitai plancommand to generate an implementation plan and prompts that you can use with your favorite coding assistant.🧪 Issue enrichment is currently in open beta.
You can configure auto-planning by selecting labels in the issue_enrichment configuration.
To disable automatic issue enrichment, add the following to your
.coderabbit.yaml:💬 Have feedback or questions? Drop into our discord!
chore: agent-claim #678 nonce=9fcacd7ad80a16c526696e63e5ad57c0
ci: enforce the first-party skill authoring rules (#681)