The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses local files in the user's home directory, specifically ~/.config/pstack/pstack-models.md for model configurations and ~/.local/state/pstack/checklists/ for task management.
These paths are used according to the skill's documented purpose for maintaining state and configuration across sessions.
There are no network operations or exfiltration patterns detected that would suggest these files are being sent externally.
[INDIRECT_PROMPT_INJECTION]: The skill processes external repository code through the grounding phase (Phase A) and the 'how' skill.
Ingestion points: Repository subsystems and existing code files.
Boundary markers: The instructions do not explicitly define delimiters for external content.
Capability inventory: The skill can read files, write to local state files, and invoke other skills in the pstack ecosystem.
Sanitization: No specific sanitization of input files is mentioned.
While this creates an attack surface common to all repository-aware coding agents, the risk is inherent to the skill's primary function of architectural analysis.

architect