The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions mandate the automatic invocation of an external command (validator-issue --auto-file) when a bug is detected with high confidence. This process transmits a diagnosis summary—derived from local log files and project configurations—to an external bug tracker without requiring user review or confirmation. This creates a path for sensitive data (such as API keys, path structures, or internal code snippets often found in debug logs) to be exfiltrated.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute various CLI commands (agent-validate list, agent-validate health, etc.). While these commands are intended for diagnostic purposes, the use of a general-purpose shell tool to execute external binaries based on AI-generated reasoning increases the attack surface if the agent's logic is manipulated.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because its primary function is to ingest and interpret data from potentially untrusted files, such as linter output, test logs, and event logs (check_*.log, review_*.json, .debug.log).
Ingestion points: The skill reads multiple files within the project's log_dir and the .validator/config.yml file.
Boundary markers: There are no instructions or delimiters provided to the agent to treat the contents of these logs as untrusted data or to ignore embedded instructions.
Capability inventory: The agent has access to Bash for command execution and the ability to send data externally via the bug-filing mechanism.
Sanitization: The skill lacks any logic to sanitize or filter the log content before processing it or including it in diagnostic summaries.

validator-help