The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes node to execute its internal telemetry and logging script (scripts/review-run-log.mjs) and utilizes various git commands for fetching, merging, committing, and pushing changes to the repository.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests data from user-provided prompts and external repository skills (e.g., address-review-bots) to generate instructions for a reviewer cohort. While it includes safety boundaries for sub-agents, the main agent performs workspace mutations based on the sub-agents' outputs.\n- [DATA_EXFILTRATION]: The skill reads and writes telemetry data, finding IDs, and session diagnostics to the ~/.codex directory. This includes logic to discover and parse historical session transcripts from other agent runs to recover results or collect usage metrics, potentially exposing sensitive metadata across sessions.

review-fix-address-bots