The Agent Skills Directory

SAFE (SAFE): The skill defines a structured workflow for performing security audits. Its instructions are consistent with its stated purpose of assisting in vulnerability assessments and threat modeling.
Indirect Prompt Injection (SAFE): The skill contains a mechanism to read project-specific learnings from a local file (.specweave/skill-memories/security.md). While this is a data ingestion point, it is a standard pattern for persistent agent memory.
Ingestion points: .specweave/skill-memories/security.md (read via cat command).
Boundary markers: Absent; the content of the memory file is treated as direct context.
Capability inventory: Bash, Read, Grep tools are available to the agent.
Sanitization: Absent; the skill relies on the integrity of the local filesystem and the .specweave directory.
Command Execution (SAFE): The skill utilizes Bash to check for local memory files. The command is hardcoded and limited to reading a specific file path within the project structure, posing no risk of arbitrary execution or privilege escalation.

security