The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes dynamic context injection (the !command syntax) to execute ls commands at load time. These commands are used to verify the state of the project's work plans and task files within the docs/plans/ directory. The skill also performs git commit operations as part of its mandatory execution cycle. These commands are legitimate for a development automation tool.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted data from project documents to influence sub-agent behavior.
Ingestion points: Reads content from files in docs/plans/ and docs/design/ (SKILL.md).
Boundary markers: Absent when reading documentation into the orchestrator's context, though a static [SYSTEM CONSTRAINT] suffix is appended to delegated sub-agent prompts.
Capability inventory: File system listing, file modification (via sub-agents), and git repository commits.
Sanitization: No validation or escaping is performed on the content of the processed documents before they are interpolated into the workflow logic.

recipe-build