Skill-Doctor: Diagnoses ineffective skills using local conversation history
aigclink · x · 2026-08-29
Skill-Doctor is a "meta-skill" that grades agent skills by analyzing real conversation history on your local machine across efficiency and code quality rubrics.
Core Logic:
- Only sessions with poor scores trigger diagnostic fixes; generic best practices are banned.
- Example: If a skill never triggered in a failed session, the issue is diagnosed as a "trigger description" problem.
Workflow:
- Verify the harness.
- Select scope (repo/all/specific projects).
- Collect data (defaults to last 45 days, max 12 sessions).
- Scoring (Efficiency, Code Quality, Coverage).
- Aggregate results and draft concrete skill edit suggestions.
Scoring Formula:
Overall = 0.5×Efficiency + 0.35×Code Quality + 0.15×Skill Coverage
Related event: Skill-Doctor Diagnoses Ineffective Agent Skills from Local Chat History(2 posts)→
More from coding & agent
- Handling regression testing for AI-written code: QA agents outperform pure unit tests — Specialist_Agent3599 · 2026-08-29
- OpenClaw's Rise and Fall: A Viral AI Agent Fades Within Months — 量子位 · 2026-08-29
- Experiment: Giving coding agents fully deterministic memory without vector DBs — madsthines · 2026-08-29
- Interop notes: Shipping OAuth on an MCP server for Claude and ChatGPT — NoStrawberry1162 · 2026-08-29
- rednote explores open-weight multimodal model with 512K context for long-horizon agents — aftahi_ai · 2026-08-29
- Dev voice tokenbender slams speed cult: fast means shipping what matters, not more crap — tokenbender · 2026-08-29