The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts included in the package for data processing tasks. In SKILL.md, the agent is directed to use fingerprint_spec.py for SHA256 hashing and envelope_artifact.py for packaging content into length-delimited artifacts. These scripts do not perform network operations or access sensitive system credentials, serving solely for data integrity and formatting.
[PROMPT_INJECTION]: As the skill synthesizes conversation history into documentation, it possesses an indirect prompt injection surface. Ingestion points: Reads from the current conversation and named repository authorities. Boundary markers: The skill defines and uses a custom artifact envelope format (---BEGIN CODEX CONVERSATION ARTIFACT---) to separate metadata from content and ensure traceability. Capability inventory: Includes local file reads, stdout writes, and execution of internal utility scripts via the Python interpreter. Sanitization: The provided Python scripts validate that all input is valid UTF-8 and strictly enforce a 10 MiB size limit to prevent resource exhaustion.

to-spec