Generate SKILL.md drafts from pipeline lessons and merged pull-request patterns.
npx skills add https://github.com/CleanExpo/Pi-Dev-Ops --skill meta-curator
Please help me install this Agent Skill. Skill: meta-curator Source: https://github.com/CleanExpo/Pi-Dev-Ops/tree/main/skills/meta-curator Command: npx skills add https://github.com/CleanExpo/Pi-Dev-Ops --skill meta-curator
Meta-curator turns the Pi-CEO engineering system’s accumulated lessons and merged-PR patterns into reusable skill drafts so the skill registry compounds over time instead of staying manually authored.
.harness/lessons.jsonl (weekly) and merged-PR diffs (daily) to generate candidate SKILL.md drafts grounded in observed workflows.Pi-Dev-Ops/skills/<proposed-name>/SKILL.md only after the user approves with 👍.Ask an AI agent to run the curator now (via the /curator:run-now command) and then review the proposed SKILL.md drafts surfaced for your approval.
High-intent search queries and answers about installing and using this skill.
Automatically generating GitHub skills from merged pull-request patterns involves clustering daily PR diffs and weekly pipeline lessons into candidate SKILL.md drafts. The system converts recurring engineering workflows into proposed skill registry entries grounded in observed evidence.
Evidence clustering for automated skill discovery groups recurring pipeline lessons and merged-PR patterns into topic or file-based clusters. This mechanism prevents duplicate skill proposals by matching new clusters against existing skills before generating drafts.
HITL approval gating routes each proposed SKILL.md through a review chat flow where the user must explicitly approve with a 👍 reaction. The draft is only written to the skill registry after this human-in-the-loop consent is verified.
Preventing duplicate skill proposals from recurring pipeline lessons is handled through cluster-based deduplication. The system matches evidence clusters against the existing skill registry, optionally proposing amendments to current skills instead of creating duplicates.
Audit trails and rate limiting are required for automated skill authoring to maintain operational safety. The system persists proposal records to curator JSONL logs, archives rejected clusters, and enforces cooldown windows to prevent re-proposing rejected evidence.
Automated skill proposals get rejected in the review loop when evidence clusters fail HITL approval gating or match existing skills. Rejected or expired clusters are archived with loop guards to prevent re-proposal within the configured cooldown window.
5hdanielCluster phase execution issues into actionable patterns and guardrail patches.
epieczkoValidate and register hook manifest files into the Hook Registry.
Agile-VAutomates foundational governance and context engineering for Agile V agents.
OlatisunkanmiDraft and validate SKILL.md files with guided prompts and repository wiring steps.