# Skill Lifecycle Promotion

Use when adding or auditing Agentlas skill lifecycle metadata, skill-registry.json...

- Repository: https://github.com/agentlas-ai/Agentlas-OS (1593 stars)
- Category: Engineering
- License: Apache-2.0
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add agentlas-ai/Agentlas-OS --skill skill-lifecycle-promotion -g -a claude-code -y`
Codex: `npx skills add agentlas-ai/Agentlas-OS --skill skill-lifecycle-promotion -g -a codex -y`
Cursor: `npx skills add agentlas-ai/Agentlas-OS --skill skill-lifecycle-promotion -g -a cursor -y`
Gemini CLI: `npx skills add agentlas-ai/Agentlas-OS --skill skill-lifecycle-promotion -g -a gemini-cli -y`
Claude Code plugin marketplace: `/plugin marketplace add agentlas-ai/Agentlas-OS`

## Description

Use when adding or auditing Agentlas skill lifecycle metadata, skill-registry.json, trial evidence, Curator promotion decisions, or first-class skill promotion gates.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Skill Lifecycle Promotion

Use this skill when a generated or packaged Agentlas repo needs governed skill
promotion metadata.

## Procedure

1. Add `.agentlas/skill-registry.json` as an export-only candidate registry.
2. Add empty `.agentlas/skill-trials.jsonl` and
   `.agentlas/curator-decisions.jsonl`.
3. Keep every skill at `tier: candidate` on export.
4. Keep `runtimeFirstClassRecallEnabled: false` until semantic curation and the
   trusted owner approval service accept the exact staged instruction/skill diff.
   Curator decisions never authorize executable file changes by themselves.
5. Add success predicates and situation tags for every skill.
6. Separate `## Memory Events` from `## Skill Trial Events`.
7. Treat LLM rubric review as weak evidence only.
8. Block promotion when authority separation, sealed holdouts, replayability, or
   rollback evidence is missing.
9. Include false-accept, blind-spot, and drift terms in any durable-error budget.

## Output

Return:

- registry files added or checked;
- promotion tier status;
- evidence gaps;
- rollback/quarantine status;
- residual risks.