lifecycle-canary-convergence — Claude Code skill: install
lifecycle-canary-convergence is a Claude Code skill that checks Memory lifecycle state through an inventory, dry-run, apply, and audit. It runs before another retry or canary.
Apache-2.0 ★ 104 updated 2026-11-17
npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a claude-codeWhat lifecycle-canary-convergence does
The skill builds a bounded snapshot of Memory lifecycle state, including durable runs and jobs, Plans and mutations, Support records, source projections, Reviews and Findings, and the vector outbox. It then moves through a report-only inventory, an exact dry-run, an authorized apply, and a strict incremental audit.
Before changing code or data, define a testable hypothesis and the evidence that would confirm or reject it. The skill separates historical data, migration residue, and one-off provider or runtime failures from reproducible contract violations. It preserves Plans, Relations, Reviews, Findings, and failed jobs; it does not invent missing evidence or rewrite history. The sequence stops when the incremental audit is clean and relevant state has not changed.
What it helps with
- inspect lifecycle state before another retry
- correlate Plans, Support records, and outbox state
- separate historical anomalies from recurring contract violations
- audit changes without rewriting lifecycle history
Example prompts for the agent
- Inventory Memory lifecycle state before another retry.
- Correlate Plans, Support records, Reviews, and outbox state.
- Classify historical lifecycle anomalies and run an incremental audit.
How to install lifecycle-canary-convergence
1Skills CLI (Vercel)
npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a claude-code2Manually: put the files in place
~/.claude/skills/lifecycle-canary-convergence/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/lifecycle-canary-convergence/
1Skills CLI (Vercel)
npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a codex2Manually: put the files in place
.agents/skills/lifecycle-canary-convergence/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a cursor2Manually: put the files in place
.cursor/skills/lifecycle-canary-convergence/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install dodo-labs-ai/mem-forge .agents/skills/lifecycle-canary-convergence2Manually: put the files in place
.github/skills/lifecycle-canary-convergence/SKILL.md1Skills CLI (Vercel)
npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a gemini-cli2Platform CLI
gemini skills install https://github.com/dodo-labs-ai/mem-forge --path .agents/skills/lifecycle-canary-convergenceManually: put the files in place
.devin/skills/lifecycle-canary-convergence/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about lifecycle-canary-convergence
How do I install lifecycle-canary-convergence in Claude Code?
Run npx skills add dodo-labs-ai/mem-forge --skill lifecycle-canary-convergence -a claude-code. The skill file is at ~/.claude/skills/lifecycle-canary-convergence/SKILL.md.
What does lifecycle-canary-convergence do?
It gathers a bounded lifecycle snapshot, then guides a report-only inventory, dry-run, authorized apply, and incremental audit.
How does the skill handle historical anomalies?
It distinguishes historical data, migration residue, and one-off failures from reproducible contract violations. Existing Plans, Relations, Reviews, Findings, and failed jobs remain intact.
Source text
SKILL.md · Apache-2.0
--- name: lifecycle-canary-convergence description: "Converge a Memory lifecycle canary, retry or data repair with one bounded inventory, dry-run, apply and audit. Use before another lifecycle retry or canary, when auditing Plans, Support, Reviews and outbox state, or when classifying historical lifecycle anomalies." --- # Lifecycle Canary Convergence - Before another retry or canary, review the complete design contract and static call path for the affected state transition: transaction ownership, operation ordering, idempotency, stale guards, retry semantics, and SQLite/Cloud adapter parity. - Define a falsifiable hypothesis and the exact confirming or rejecting evidence before changing code or data. Classify historical data, migration residue, and one-off provider/runtime failures separately from reproducible shared-contract violations. - Prefer one bounded joint snapshot that correlates durable runs/jobs, Lifecycle Plans and mutations, current and stale Support, Source Projection lineage, Reviews/Findings, vector outbox, and process health. Do not replace it with repeated narrow probes that guess at adjacent causes. - Use a bounded convergence sequence: one report-only inventory, one exact dry-run, one controlled apply when authorized, and one strict incremental audit. Stop when the incremental audit is clean and no relevant state changed. - Never fabricate Evidence, infer missing provenance from semantic similarity, or silently rewrite lifecycle history to make an audit pass. Preserve historical Plans, Relations, Reviews, Findings, and failed jobs. - Report current and historical populations separately. Distinguish active Memories from superseded or retired lifecycle records and classify unexpected growth by lifecycle reason and bounded time cohort.
© dodo-labs-ai · Apache-2.0 · dodo-labs-ai/mem-forge