The Agent Skills Directory

[SAFE]: The skill demonstrates legitimate orchestration behavior, including reading configuration files and managing track data within the local directory structure. All operations are consistent with its described purpose.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it transforms unvetted user goals from $ARGUMENTS into project plans and directory paths. 1. Ingestion points: User goal input via $ARGUMENTS in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Local file and directory creation in SKILL.md and invocation of internal agents. 4. Sanitization: No explicit sanitization of input goal strings is documented before their use in planning or file path construction. This surface is characteristic of goal-driven orchestrators and does not represent malicious intent.

go