The Agent Skills Directory

[SAFE]: The skill's functionality is limited to reading and displaying content from a local directory, which is a standard pattern for project-specific documentation management.
[COMMAND_EXECUTION]: Uses the Bash tool for a simple directory check and the Grep tool for searching text within the loaded files. These operations are constrained to the local environment and the skill's stated purpose.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it ingests data from local files and incorporates it into the agent's context. 1. Ingestion points: Files located in the .workflow/specs/ directory (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Uses the Read, Bash, and Grep tools (SKILL.md). 4. Sanitization: No specific sanitization or escaping of the ingested file content is mentioned.

spec-load