The Agent Skills Directory

[SAFE]: The skill restricts all file operations to the docs/threads/ directory, specifically managing documentation files like ledger.md and proposal.md. This localized scope prevents unauthorized access to sensitive system or user data.
[NO_CODE]: The skill consists entirely of natural language instructions and markdown templates without any executable scripts or binary files, eliminating the risk of direct malicious code execution.
[PROMPT_INJECTION]: The skill has an ingestion surface for indirect prompt injection as it processes user prompts and referenced artifacts to generate documentation. However, the risk is minimal because the agent's capabilities are restricted to writing text files; it lacks network access, shell command execution, and the ability to perform automatic commits.
Ingestion points: The skill reads 'referenced input' and ledger.md from the docs/threads/ directory.
Boundary markers: No explicit boundary markers or 'ignore' instructions are used for external input.
Capability inventory: The skill permits reading and writing markdown files in project directories.
Sanitization: No content filtering or escaping is implemented for the ingested text.
[SAFE]: The skill includes an explicit policy that forbids automatic commits, ensuring that all changes made by the agent must be reviewed and handled by the user or an external orchestration process.

propose