The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes agent-validate commands via the Bash tool. These commands are used to clean, run, and check the status of validation tasks. This tool appears to be a legitimate utility provided by the vendor (Codagent-AI).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its error extraction and fixing logic (Step 5 and Appendix). It parses external log files located in validator_logs/ for specific patterns such as --- Fix Instructions --- and --- Fix Skill: <name> ---, then directs the agent to execute those instructions or skills. If a file being validated by the system contains malicious code designed to trigger specific validator output, an attacker could potentially influence the agent's subsequent actions.
Ingestion points: Log files in the validator_logs/ directory (referenced in Step 3, Step 4, and the Extract Prompt).
Boundary markers: Absent; the skill does not include instructions to ignore or sanitize embedded instructions within the logs.
Capability inventory: The skill has access to the Bash and Task tools, and the logic allows for the invocation of other skills.
Sanitization: Absent; the extracted text is used directly to guide the agent's next steps.

validator-check