The Agent Skills Directory

[SAFE]: The skill implements a security-conscious installation check at the start of its execution, verifying that all referenced internal knowledge notes are present before proceeding. This prevents the agent from operating on incomplete or potentially corrupted context.
[SAFE]: The allowed-tools configuration uses specific command-level scoping for Bash (restricting it to cat, ls, and find), which adheres to the principle of least privilege and prevents arbitrary command execution.
[SAFE]: While the skill reads local project files such as CONTRIBUTING.md and issue templates, this is done solely to analyze existing processes for documentation updates. There is no evidence of data exfiltration or sensitive information harvesting.
[SAFE]: The skill's instructions focus on generating documentation and structured templates (like GitHub Issue Forms) and do not contain any prompt injection attempts, obfuscation, or persistence mechanisms.

contribution-workflow