The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the agent-validate command-line utility via the Bash tool to perform cleanup, validation checks, and log management.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes data from external log files to determine its next actions.
Ingestion points: The agent is instructed to read contents from the validator_logs/ directory and specific log files such as console.N.log or individual check logs.
Boundary markers: The skill relies on plaintext markers like --- Fix Instructions --- and --- Fix Skill: <name> --- to identify instructions. These markers could be spoofed if an attacker is able to influence the content of the logs, for example, by placing malicious strings within source code that appear in error messages.
Capability inventory: The agent has access to the Bash tool and the Task subagent tool, providing a path to execute commands based on the ingested log data.
Sanitization: No sanitization or validation of the extracted 'Fix Instructions' is specified before the agent follows or executes them.

validator-check