The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (../squad/scripts/observe.py) and shell utilities such as jq and mktemp to manage data and telemetry. These operations are performed using relative paths within the project's directory structure and are integral to the skill's workflow.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and analyzes data from external task descriptions and codebase contents.
Ingestion points: Data is fetched from the task board API (GET /task/$ID) and read directly from local files and schemas to provide context for the refinement interview.
Boundary markers: There are no explicit delimiters or safety warnings used when interpolating task descriptions into the analysis prompt.
Capability inventory: The agent has permissions to read files, perform network requests via the api tool, and execute local subprocesses.
Sanitization: The skill employs jq to construct JSON payloads for API writes, which provides appropriate escaping for strings synthesized during the refinement process.

squad-refine