The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script (generate-task-graph.mjs) via a bash wrapper to process task definitions and generate output graphs. This is a legitimate part of the skill's primary functionality.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes a YAML file (tasks.yaml) which may contain content derived from untrusted user feature requests.
Ingestion points: Reads tasks.yaml and SPEC.md from the planning/ directory.
Boundary markers: The skill relies on a strict YAML schema but lacks explicit boundary markers to prevent the execution of instructions embedded within task fields like notes or context.
Capability inventory: The skill executes a JavaScript-based validator and generator script to process input files.
Sanitization: No explicit sanitization or input validation for instructions embedded in data fields was found in the provided file snippets.

feature-tasks