The Agent Skills Directory

[SAFE]: The skill's primary objective is to facilitate feature porting by creating a 'gap map' analysis. It is explicitly restricted from implementing code, modifying configurations, or auto-chaining to other tasks, ensuring its actions are confined to documentation and planning.
[SAFE]: File access is restricted to local repository context (e.g., GLOSSARY.md, ADRs, and source code) using standard search utilities like rg and git grep. It avoids bulk reading and has no network communication capabilities, preventing data exfiltration risks.
[SAFE]: The instructions emphasize integrity and verification, requiring the agent to stop and ask for user input rather than guessing and mandating that all claims in the generated artifact be backed by concrete evidence references.
[SAFE]: The 'Zero attribution' policy is a stylistic requirement for the generated output and does not interfere with safety guidelines or deceive users regarding the skill's technical capabilities.
[SAFE]: Documentation in the evaluation folder includes bash and python commands for running internal test suites. These are standard development instructions for maintainers and do not represent remote code execution vulnerabilities for the skill itself.

port-feature