The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from Requirement Markdown documents and TestPoint lists which could contain malicious instructions.
Ingestion points: Data is retrieved via cawplan knowledge documents get and cawplan qa-insights testpoints list in SKILL.md.
Boundary markers: No explicit delimiters or sanitization logic are defined for the ingested content, and the 'Source Fidelity' rule (源值保真) instructs the agent to retain source text verbatim.
Capability inventory: The skill has access to the Bash tool, allowing it to execute system commands and write files.
Sanitization: No sanitization or filtering of external content is specified before processing or inclusion in output.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute various cawplan CLI commands and a local Node.js script for data processing and export.
Evidence: Multiple invocations of cawplan and node scripts/export_to_csv.js are present in SKILL.md.
[DYNAMIC_EXECUTION]: The skill generates temporary JSON files at runtime to pass data between the agent's context and a local export script.
Evidence: SKILL.md contains logic to write session data into /tmp/a3_export_*.json before executing the export script.

cawplan-testcase-generate