The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's Python scripts, specifically launch_codex_worker.py and run_codex_audit_loop.py, utilize the subprocess module to dynamically construct and execute commands for the codex CLI tool. These commands are built using strings and paths derived from external mode configuration files and user-provided arguments.
[PROMPT_INJECTION]: The skill implements an automated orchestration loop that exhibits an indirect prompt injection surface. The run_codex_audit_loop.py script parses an AI-generated audit report to extract 'handoff blocks,' which are then directly interpolated into prompts for subsequent workers.
Ingestion points: The orchestrator reads the AUDIT_REPORT.md file produced by the previous audit worker.
Boundary markers: Extracted handoff content is interpolated into new prompts without explicit delimiters or instructions to ignore nested instructions.
Capability inventory: The worker agents have the capability to read and write files within the repository and perform git commits via the codex tool.
Sanitization: The extracted handoff content is not validated or sanitized before being included in subsequent prompts.
[EXTERNAL_DOWNLOADS]: Orchestration templates and documentation within the skill suggest the use of npx -y codex mcp-server, which downloads and executes the Codex MCP server from the npm registry.

domain-reviewer