The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local log files and user descriptions which may contain adversarial instructions intended to influence the agent's behavior during the issue drafting process.
Ingestion points: Reads from $ARGUMENTS, user responses, .validator/config.yml, .debug.log, and .execution_state.
Boundary markers: The skill does not employ explicit delimiters or system instructions to isolate ingested data from the agent's control logic.
Capability inventory: The skill uses the Bash tool to execute tail, mktemp, and gh issue create commands.
Sanitization: There are manual instructions for the agent to redact sensitive values like tokens and API keys, but no automated validation or filtering is implemented.
[DATA_EXFILTRATION]: The skill is configured to read local files and send their contents to an external GitHub repository.
Sensitive file paths: The skill reads .validator/config.yml, a common location for configuration data that might include sensitive credentials.
Network operations: The gh issue create command is used to transmit the collected data to the Codagent-AI/agent-validator repository.
Risk factor: While the skill includes redaction steps, the --auto-file mode allows the data to be sent without a preview or explicit user confirmation, increasing the risk of accidental data exposure.

validator-issue