The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests data from external iteration artifacts to update project documentation.
Ingestion points: The skill reads .agents/flow/it_{iteration}_PRD.json, .agents/flow/it_{iteration}_refactor-report.md, and .agents/flow/it_{iteration}_progress.json (SKILL.md).
Boundary markers: Absent; there are no specific instructions or delimiters provided to the agent to distinguish data from instructions within these files.
Capability inventory: The skill is capable of reading and writing to sensitive project documentation and configuration files including .agents/PROJECT_CONTEXT.md, ROADMAP.md, AGENTS.md, and README.md (SKILL.md).
Sanitization: Absent; the content from iteration artifacts is processed without explicit sanitization or validation.
[COMMAND_EXECUTION]: The skill performs automated file system operations, specifically reading project state and writing updates to core documentation files. This behavior is mitigated by a mandatory human-in-the-loop requirement where the user must approve a summary of changes before any files are modified.

approve-prototype