Skills that Improve Development loops run: review pull request history for agent context and Tessl verifier improvements, and read recent agent runs to record recurring problems as project findings.
77
97%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Survey recent PR and review activity lightly. Prefer active code areas and recent acknowledged corrections, while keeping all repository history available. Widen discovery when recent activity is thin or related evidence lies elsewhere. Use changed paths, not titles alone, to establish three distinct areas and a question for each. Different application directories may implement the same subsystem.
Select questions about decisions agents could repeatedly get wrong in ordinary code changes. Do not restrict discovery to AGENTS.md or skill titles, or make this improvement skill's own follow-up PRs the entire investigation. Prioritize implementation present at the evidence revision; replace entirely absent branch-only areas when useful alternatives exist.
Do not inventory every PR or file in a date range before investigating. Save discovery queries, sampled PR metadata, available counts and unfinished pagination. A few updated closed-PR pages are a sample, not complete merge history. Reconcile unique results against reported counts and account for search caps only when claiming a query is exhausted.