The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with local CLI tools. It executes plannotator annotate <path> --gate --json for plan review and uses do-todo to update task lists. These operations are core to the skill's stated purpose of plan implementation and verification.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from requirements files and feature descriptions to generate structured plans. While it includes instructions to avoid inventing product behavior, it lacks explicit data sanitization.
Ingestion points: Reads requirements files and feature descriptions in the Intake section of SKILL.md.
Boundary markers: The skill uses sections and specific headers but does not explicitly wrap external content in security-focused delimiters.
Capability inventory: Writes markdown files to the local filesystem (e.g., docs/plans/), executes the plannotator CLI, and calls the do-todo tool.
Sanitization: No explicit sanitization or escaping of external requirements text is defined before it is incorporated into the generated plan.

do-plan