The Agent Skills Directory

[SAFE]: The skill instructions define a structured workflow for product management tasks, focusing on requirements gathering, user stories, and acceptance criteria.
[DATA_EXPOSURE]: The skill accesses project-specific directories and memory files located within ~/.ai-software-agency/. These paths are consistent with the skill's operational context and do not attempt to access sensitive system-level credentials such as SSH keys or cloud provider configurations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including a goal brief and user-provided answers to clarifying questions. While this creates a surface for indirect prompt injection, it is managed through a documented governance process. \n
Ingestion points: Discovery Phase (goal brief, goals.md, and user answers) and specialist output files. \n
Boundary markers: Absent. \n
Capability inventory: Write, Edit, TaskCreate, and Agent (sub-agent spawning) capabilities across the workflow. \n
Sanitization: None explicitly stated; however, the workflow mandates a manual approval gate (GATE_READY to CEO) before completion, providing human-in-the-loop oversight.

product-manager