The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs file system operations within the local project directory.
Evidence:
Reads workflow state from .pic/state.json.
Enumerates existing files in .pic/integration-results/ to generate unique IDs.
Writes detailed test reports to .pic/integration-results/INT-[NNN].md.
Appends event logs to .pic/status-log.jsonl.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface where user-provided input and generated content are interpolated into a sub-agent prompt.
Ingestion points: The $ARGS input (components to test) and the Test Plan generated in Step 4.
Boundary markers: Absent. External data is directly interpolated into the prompt for the test-writer sub-agent without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill and its sub-agents have the capability to read and write files within the project's .pic/ directory and use the Task tool.
Sanitization: Absent. No filtering or escaping is applied to the component list or test plan before interpolation.

pic-integration