The Agent Skills Directory

[SAFE]: The skill establishes a comprehensive debugging framework with a focus on evidence and root-cause analysis. It includes proactive security measures such as a mandatory redaction gate for shared memory and explicit prohibitions against logging PII or credentials.
[COMMAND_EXECUTION]: The skill utilizes standard, non-privileged development tools like git, grep, and system performance utilities (iostat, vm_stat, top) for legitimate diagnostic tasks. Safety protocols are included to prevent accidental repository corruption, such as checking worktree status before starting a bisect operation.
[INDIRECT_PROMPT_INJECTION]: Although the skill is designed to ingest potentially untrusted external data like logs and stack traces, it mitigates injection risks by instructing the agent to treat such data as evidence to be verified rather than instructions to be followed. The methodology requires a specific, testable hypothesis and independent verification before any code changes occur.

hunt