test-collaboration 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
• SKILL.md references templates/TESTS.example.md, but the submitted skill directory only contains SKILL.md; the referenced template is missing or was not included.
• Indices de permissions à haut risque : exécution shell ou de commande
• The workflow may run test commands to verify exit codes, but it does not explicitly instruct the agent to review commands before execution or avoid running unfamiliar/untested commands outside a safe environment.
• The skill reads Spec/Issue/Bug files that could contain untrusted or injected instructions, but it does not explicitly state that those files are data and must not override the skill's own governance rules.