Capture learning
Private, local-first second brain for AI agents, with shared memory and governed skills across Claude Code, Codex, and your Tailnet.
npx -y skills add Chulinuwu/skillloom --skill capture-learningAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 21 days oldThe repository was created 21 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Capture a reusable workflow as a Skillloom-managed Agent Skill candidate with provenance, validation, trust scanning, approval, promotion, and rollback evidence. Use after a complex task reveals repeatable steps, when the user asks to learn or save a workflow as a skill, or when verified task evidence should become a portable skill. Do not use for one-off notes or autonomous promotion.
SKILL.md
2.3 KB, 379 tokens by cl100k_base, as published. Nobody here has run it
Capture Learning
Turn verified task evidence into one focused, harness-neutral Agent Skill candidate. Treat drafting and promotion as separate decisions. Respond in the language of the user's latest message. Keep exact identifiers, hashes, paths, and source titles unchanged.
Workflow
- Confirm that the workflow is reusable and identify the narrow behavior the skill should preserve.
- Create a temporary skill directory containing
SKILL.mdand only the resources required by that workflow. Exclude credentials, full transcripts, unrelated artifacts, symlinks, binaries, and machine-specific instructions. - Record concise provenance such as a task or issue identifier. Do not store conversation bodies as evidence.
- Run
skillloom capture <skill-directory> --created-by agent --evidence <reference>and retain the candidate and operation identifiers. - Run
skillloom validate <candidate-id>. Review structural validation and every trust finding. Treat the scan as preflight evidence, not proof that content is safe. - Present the candidate summary, provenance, content hash, findings, intended targets, and reload impact. Stop when a danger finding exists. Ask before accepting warnings or promoting.
- Only after explicit approval, run
skillloom promote <candidate-id> --target <targets> --scope <scope> --yes. Add--accept-warningsonly when the user explicitly accepts the reported warnings. - Verify the promotion record, destination hashes, and target results. Report the promotion identifier and whether a reload or new session is required.
- If verification fails, inspect
skillloom statusand useskillloom resume <operation-id> --yesonly when its staged evidence still agrees. Otherwise runskillloom rollback <promotion-id> --yesand report the restored hashes.
Never promote autonomously. A request to capture or validate is not approval to publish.