The Agent Skills Directory

[SAFE]: The skill includes a robust security section ('Untrusted content boundary') that explicitly instructs the agent to treat all retrieved external material as untrusted data rather than instructions. It mandates the use of delimiters, validation, and user confirmation for high-risk actions like credential use or command execution.
[SAFE]: The 'Freshness check' mechanism defines specific triggers (cross-session work, external API usage, or protected branch work) where the agent must verify the current state of the environment using standard tools like git log, grep, and curl before proceeding.
[SAFE]: The instructions implement a specific safety guardrail for integration work, requiring explicit human consent before any changes are applied to the 'main' or 'master' branches.
[SAFE]: The skill references other components in a logical development ecosystem (e.g., git-worktrees, writing-plans, finishing-a-branch) to maintain workspace isolation and process integrity.

executing-plans