The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the git status --short command during its startup scan to identify existing changes and inform its planning logic. This is a standard read-only operation for repository-aware agents.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific documentation and configuration files from the repository to guide its planning and auditing tasks, which constitutes a potential surface for indirect prompt injection.
Ingestion points: The 'Startup scan' phase reads local files such as .codestable/INDEX.md, .codestable/attention.md, and .codestable/reference/project-knowledge-contract.md.
Boundary markers: The instructions do not specify explicit delimiters or 'ignore' instructions to isolate content read from files from the agent's core instructions.
Capability inventory: The skill's actions are restricted to reading repository files, executing git status, and writing planning artifacts (e.g., context packs and task memory) to the .codestable/tasks/ directory. It is explicitly instructed not to modify source code during audit operations.
Sanitization: No explicit sanitization or validation of the ingested file content is mentioned.

cs-plan