The Agent Skills Directory

[SAFE]: The skill is designed for documentation and workflow management with no detected malicious intent or patterns.
[COMMAND_EXECUTION]: The skill configuration limits shell tool usage to git commands (Bash(git:*)), significantly reducing the risk of unauthorized system command execution.
[DATA_EXPOSURE]: The skill only interacts with standard documentation paths like specs/ or docs/specs/ and does not attempt to access sensitive system files or credentials.
[PROMPT_INJECTION]: No patterns of instruction overriding, DAN-style prompts, or safety filter bypasses were detected in the skill instructions or guides.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific data (specification files) which is a potential injection surface. However, the risk is negligible as it lacks high-privilege capabilities and performs standard documentation tasks.
Ingestion points: Files within docs/specs/, specs/, openspec/changes/, and .specify/ (referenced in guide.md).
Boundary markers: None explicitly defined in the logic for the agent to distinguish between its instructions and the content of the specs.
Capability inventory: Limited to Read, Write, Grep, Glob, and Bash(git:*) across both SKILL.md and guide.md.
Sanitization: No explicit sanitization or validation of the processed document content is mentioned.

sdd