← ClaudeAtlas

evolve-applylisted

Applies a human-approved /evolve proposal. Runs evolve-apply.sh to patch core/, convert, and audit; reverts and records a failure ledger if the audit fails. NEVER run without explicit human approval of the proposal. <example> Context: Human reviewed proposal abc12345 and approved it. user: "/evolve-apply abc12345" assistant: "[EVOLVE-APPLY] abc12345 applied — convert + audit green." </example>
reghis86/context-forge · ★ 0 · AI & Automation · score 71
Install: claude install-skill reghis86/context-forge
## What This Does Applies one approved proposal from `<IDE_DIR>/evolve/pending/<id>.yaml` to the source tree under the human gate. This is the ONLY path by which the self-evolving loop mutates `core/`. ## How to Use 1. Confirm the human has explicitly approved `<id>` (NEVER infer approval). 2. Run: `bash core/scripts/tools/evolve-apply.sh <id>`. 3. On `[EVOLVE-APPLY] <id> applied`: the patch landed, `npm run convert` and `npm run test:audit` passed, and the proposal moved to `applied/`. Commit it. 4. On `reverted`: the patch or the audit failed; the tree is clean. Invoke `/record-failure` to log a ledger, then report the failure. Do NOT retry blindly. Every invocation appends one row to `<IDE_DIR>/evolve/ledger.tsv` (`applied | reverted | reverted-patch`) — the persistent outcome record `/evolve` reads Tier-0 to avoid re-proposing patches that were already reverted. ## Constraints NEVER run without explicit human approval of the specific proposal id. NEVER edit the patch_text — apply it verbatim or report failure. ALWAYS rely on evolve-apply.sh to revert on audit failure (do not hand-patch). NEVER apply more than one proposal per invocation. ALWAYS commit applied changes so the audit trail (git) stays intact.