The Agent Skills Directory

[SAFE]: The skill implements significant safety controls including mandatory execution budgets (max_loop_iters, max_repair_rounds) and Human-In-The-Loop (HITL) gates for high-stakes or destructive operations.\n- [COMMAND_EXECUTION]: A Node.js validation script (scripts/validate-skill.mjs) is provided to check the skill's structural integrity. Analysis confirms the script only performs read-only filesystem operations and string comparisons on the skill's own files.\n- [DATA_EXFILTRATION]: No evidence of unauthorized network access, credential harvesting, or data exfiltration was found. The skill uses local filesystem paths for configuration.\n- [EXTERNAL_DOWNLOADS]: The skill does not perform any remote downloads or fetch external code. All components are self-contained or reference local files.

control-graph