read-book Rapport d’évaluation.

Plan de validation

Ce que l’Agent doit faire ensuite

1. 1Inspect repository, README/SKILL.md, license, and recent commits before production use.
2. 2Install in an isolated workspace or sandbox with no production secrets available.
3. 3Run the smallest representative task and record files touched, commands run, network access, and outputs.
4. 4Compare the selected skill against at least one alternative when the eval status is review or failed.
5. 5Promote only after the agent reports a successful verification result and unresolved warnings are accepted.

Ne pas utiliser quand

Conditions exigeant une autre skill

• Équipes qui nécessitent un SLA soutenu par le fournisseur
• production agents without a repository review
• No critical security risks found. The skill writes to local book-note directories and only writes to the second-brain vault after explicit user confirmation.
• Minor security consideration: file paths and book metadata-derived slugs should be sanitized before use in shell commands or filesystem paths to avoid path traversal or command injection from untrusted book metadata.
• Quality score needs review
• Production credentials, payments, or irreversible account changes without explicit human review