# Cockpit Skillops > Run SkillOps (debrief -> distill -> lint) so cockpit skills continuously learn from real task outcomes. Use when this capability is needed. - Skill: `tomevault-io/cockpit-skillops` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/cockpit-skillops` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/cockpit-skillops/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Productivity - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/cockpit-skillops --- # Cockpit SkillOps ## When to use - After completing a meaningful task or slice. - Before reporting final `outcome="done"` for autopilot tasks that require SkillOps evidence. ## Required sequence 1. `node scripts/skillops.mjs debrief --skills --title "What changed"` Fast path when the learning is already clear: `node scripts/skillops.mjs debrief --skills --skill-update "skill-a:1-line rule" --title "What changed"` 2. `node scripts/skillops.mjs distill` 3. `node scripts/skillops.mjs lint` ## Evidence contract - Include all three commands in `testsToRun`. - Include the debrief log markdown path in `artifacts` (under `.codex/skill-ops/logs/...`). - `distill` is non-durable. Do not claim it patched durable skill files. - Raw SkillOps logs are local evidence only. - `queued` logs are non-blocking runtime evidence, not removable dirt; runtime retires them only after processed mark-back succeeds. - Do not pretend a learned heuristic is durable until runtime either: - retires the empty log locally, or - queues the dedicated `skillops-promotion` lane and later promotes the heuristic on the promotion PR branch. ## Learned heuristics (SkillOps) --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/future3ooo) — claim your Tome and manage your conversions.