The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Python subprocess module to interface with the codex command-line utility and standard system tools such as which. This execution is confined to the specific binary and is essential for the skill's primary function as an interoperability bridge.
[INDIRECT_PROMPT_INJECTION]: The skill reads project-level metadata and configuration from CLAUDE.md and various skill definition files to generate AGENTS.md. This creates an indirect prompt injection surface where malicious instructions embedded in a project's documentation could be propagated to the generated documentation. However, this is a common characteristic of documentation tools and not an indicator of malicious intent within the skill itself.
Ingestion points: claude_parser.py reads CLAUDE.md, .claude/skills/, and .claude/agents/.
Boundary markers: The generated AGENTS.md does not use explicit boundary delimiters for included project content.
Capability inventory: Provides shell command execution via the codex CLI through codex_executor.py.
Sanitization: Content extracted from the project is included in the output without specific sanitization or escaping.
[DATA_EXPOSURE]: The skill accesses project-specific files and the standard Claude Code user configuration directory (~/.claude/skills) to discover project assets. It does not attempt to access sensitive system files, credentials, or environment variables.

codex-cli-bridge