The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to build and execute shell commands using codex exec. It pipes potentially untrusted content (plans, code) directly into the echo command without explicit escaping or sanitization.- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to malicious instructions embedded in the code or plans it processes. Ingestion points: Phase 2 and Phase 5 in SKILL.md ingest architecture plans and implementation code. Boundary markers: The skill uses simple double-quote delimiters within a shell command template, which is insufficient to prevent context breakout. Capability inventory: The skill utilizes codex exec for tool execution and provides access to Edit, Write, and Read tools for file system interaction across all scripts. Sanitization: There are no instructions for escaping, validating, or filtering external content before interpolation.- [DYNAMIC_EXECUTION]: The skill dynamically assembles and executes shell commands at runtime based on user-supplied parameters and agent-generated content.

codex-claude-loop