The Agent Skills Directory

[SAFE]: The skill is designed for local file management and transformation within a specific workspace (vault/). It follows a strict set of prohibitions, such as never modifying raw source files and only writing to a designated drafts directory.
[COMMAND_EXECUTION]: The skill uses Bash restricted to git status and git log. These are used for monitoring the repository state and are considered safe, read-only operations in this context.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data from raw files. However, the risk is mitigated by the architectural design.
Ingestion points: Reads files from vault/01_raw/ using the Read tool in Step 1.
Boundary markers: No explicit delimiters are used to separate raw content from agent instructions.
Capability inventory: Access to Read, Grep, Glob, Write, and restricted Bash commands.
Sanitization: No explicit sanitization of the input text is described.
Mitigation: The risk is effectively controlled by two mandatory 'Discussion Checkpoints' (Step 3 and Step 6) where the user must manually review and approve the extracted topics and the final draft content before any file is written to the system.

ingest