The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were detected. The skill's operations, including reading and writing to the .rcs/ directory and invoking other internal skills, are consistent with its primary purpose of strategic planning and architectural review.
[PROMPT_INJECTION]: While the skill exhibits an attack surface for indirect prompt injection due to its ingestion of codebase data and existing plan files, this is a necessary functional component of a planning tool. The skill handles this safely by separating the planning context from the execution context.
Ingestion points: SKILL.md (Review mode reads from .rcs/plans/; Interview mode gathers codebase facts via an explore agent).
Boundary markers: Absent (Standard delimiters for codebase facts are used by the underlying platform, but not explicitly defined in this skill's instructions).
Capability inventory: SKILL.md (Writes to .rcs/plans/; triggers implementation via $forge or $team).
Sanitization: Absent (Relies on the specialized execution skills to validate inputs before proceeding with code changes).

plan