The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized network operations were detected. The skill operates locally within the .local/specs/ directory.- [PROMPT_INJECTION]: The skill scans project files and accepts user input to generate technical plans and implementation tasks. While this creates an indirect prompt injection surface (Ingestion points: project files in Phase: init and Phase 2: plan; Capability inventory: file system read/write; Sanitization: none explicitly described), the risk is mitigated by explicit boundary markers and instructions. The skill mandates using [NEEDS CLARIFICATION] tags for ambiguities and includes a 'Three-Tier Boundary System' that expressly forbids the agent from committing secrets or deleting data without backups.

sdd