Diagnose and repair failing skill checks through a test-fix-retest loop.
npx skills add https://github.com/wsman/MY-DOGE-MICRO --skill skill-improve-wsman
Please help me install this Agent Skill. Skill: skill-improve Source: https://github.com/wsman/MY-DOGE-MICRO/tree/main/.agents/skills/skill-improve Command: npx skills add https://github.com/wsman/MY-DOGE-MICRO --skill skill-improve-wsman
This Skill helps refine an existing skill when it has failing checks, weak workflow coverage, or missing guardrails. It turns diagnosis and patching into a controlled loop so improvements can be measured before changes are kept.
Ask the assistant to improve a named skill using the test-fix-retest workflow and review the proposed patch before approving any write.
High-intent search queries and answers about installing and using this skill.
Run a test-fix-retest loop to diagnose static analysis failures, map them to specific gaps in the skill specification, apply patches, and verify improvements through retesting.
A test-fix-retest loop is a controlled workflow that runs baseline checks, applies targeted edits to fix failures, and retests to measure score changes before preserving or reverting modifications.
Safely patch a skill specification by proposing focused edits, asking for approval before writing, and using controlled file writes with rollback to revert changes if the retest score does not improve.
Maintain dual-domain parity by preserving game content while adding product-specific routing, steps, outputs, and follow-up guidance when a skill spans both domains during the review process.
Diagnose missing guardrails and workflow coverage by running baseline static checks and category rubric reviews, then mapping each warning to specific gaps like missing ask-before-write guidance.
Automated workflow repair requires reading the full skill specification to compare baseline and retest results, meaning changes cannot be kept automatically without confirming the score has improved.
Negentropy-LabyValidate skill definitions across static, spec, category, and audit modes.
lawless-mOrchestrate test-fail-issue-fix loops with GitHub issues and Claude Code.
onelee85Validate and repair inconsistencies in the .planning/ directory.
goupup-aiJudge whether experimental results support intended research claims.
