The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external logs which could contain malicious instructions designed to influence the agent's behavior during classification or reporting.
Ingestion points: Fetches content and assistant_message fields from cawplan session human-input-logs or product-human-input-logs (SKILL.md).
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the processed log data.
Capability inventory: The skill executes cawplan CLI commands, writes data to /tmp/humaninput_rows.jsonl, and performs classification logic (SKILL.md).
Sanitization: There is no evidence of input sanitization or validation of the text content retrieved from the logs before it is processed.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute several system commands and internal CLI tools.
Evidence: Uses cawplan, jq, and standard shell utilities (wc, echo, while loops) to manage data flow and generate reports.

cawplan-internal-qa-coding-humaninputs-test