The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted requirement data and test point descriptions fetched from an external source, which may contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Requirement links and IDs provided as arguments, as well as requirement and test point data fetched via the cawplan CLI.
Boundary markers: The skill provides instructions (Red Line 0) to avoid data invention, but does not define explicit delimiters or 'ignore' warnings for the data interpolated into prompts.
Capability inventory: The skill has access to the Bash tool, which is used to write to temporary files in /tmp/ and execute a local Node.js script.
Sanitization: The skill implements RFC4180 CSV escaping in its export_to_csv.js script to handle special characters in the exported data.
[DYNAMIC_EXECUTION]: The skill assembles a data payload at runtime and executes a bundled script to process it.
Evidence: In SKILL.md, the workflow writes a JSON representation of current test cases to a temporary file (/tmp/a3_export_*.json) and then executes node scripts/export_to_csv.js to generate the final CSV file. This is a controlled execution of a static, bundled script.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform system-level operations.
Evidence: The skill executes the cawplan CLI for fetching requirements and test points, and uses node to run its internal export script.

cawplan-testcase-generate