The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard git commands (e.g., git diff, git log, git commit, git revert) to manage debugging cycles, track code changes, and provide rollback capabilities. This is consistent with the skill's intended purpose of managing a debugging session.\n- [DATA_EXFILTRATION]: Access is restricted to project-specific paths (~/.claude/projects/, /tmp/) and standard configuration files (CLAUDE.md). No network exfiltration or unauthorized access to sensitive user credentials was observed.\n- [EXTERNAL_DOWNLOADS]: The skill does not perform any external downloads or install unverifiable dependencies at runtime. It relies on a suite of internal agent tools for its operations.\n- [PROMPT_INJECTION]: The orchestrator implements a 'Trust-but-verify' model where a synthesis agent cross-references findings from multiple independent investigators. Furthermore, it uses mandatory quality gates for both hypotheses and code fixes to mitigate the risk of accidental or adversarial misinformation from untrusted inputs.\n- [SAFE]: The skill follows established platform conventions for disk-mediated dispatch and session recovery, ensuring transparency through mandatory status updates and state persistence.

debugging