The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external specification files (SPEC_FILE) and logs (LOG_FILE) as core inputs for the implementation and inspection workflow. Maliciously crafted instructions within these external files could potentially influence the agent's behavior during the build or review phases.
Ingestion points: The skill ingests data from SPEC_FILE (mapped to the runner's --plan argument) and LOG_FILE.
Boundary markers: There are no explicit instructions to use boundary markers or to ignore embedded instructions within the ingested files.
Capability inventory: The skill possesses the capability to execute the implementation runner (build), the inspection tool (inspect), and user-defined proof checks (PROOF_CMD).
Sanitization: The instructions do not define any sanitization, validation, or filtering for the content of the external files before processing.
[DYNAMIC_EXECUTION]: The skill environment preserves and utilizes a PROOF_CMD variable, which is intended to be executed to verify changes. Executing arbitrary commands defined in project configuration or user-supplied specifications carries a risk of unintended code execution if the source is untrusted.

codex-build