The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the @openai/codex-sdk package. This is a library provided by OpenAI, which is a well-known technology service.
[COMMAND_EXECUTION]: The instructions describe the use of local Python utility scripts (e.g., scripts/codex_jsonl_to_sqlite.py) and the codex CLI tool. These operations are standard for the skill's purpose of automating developer workflows and managing execution logs.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it is designed to process untrusted external data such as repository contents and code diffs. The skill mitigates this by recommending structured output validation, sandboxing, and explicit prompt-injection defenses.
Ingestion points: The skill inventories repo roots, diffs, and command outputs as context for the agent.
Boundary markers: Documentation references specific defense guidelines in references/safety-and-execpolicy.md.
Capability inventory: The skill utilizes the codex CLI for execution and local Python scripts for database operations.
Sanitization: The instructions advocate for JSON Schema validation and sandboxing policy rules.

codex-sdk