The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple platform-specific helper binaries located in ~/.claude/skills/gstack/bin/ (e.g., gstack-skill-start, gstack-question-log, gstack-skill-end) to handle initialization, telemetry, and user preference tracking. It also runs standard development commands including xcodebuild, swift build, and nm to verify the state and integrity of the iOS project during Phase 2 and Phase 3.
[INDIRECT_PROMPT_INJECTION]: The skill reads and analyzes external application source code using Glob and Grep tools to identify debug wiring. This process creates an indirect prompt injection surface as malicious content or instructions embedded in the analyzed code could influence the agent's cleanup decisions.
Ingestion points: Phase 1 Inventory (scanning app source code via Glob and Grep).
Boundary markers: The skill requires an explicit AskUserQuestion confirmation for each identified item before proceeding with removal.
Capability inventory: The skill possesses Bash, Edit, Glob, Grep, and Read capabilities across all operational scripts.
Sanitization: Security relies on the manual user review and approval process triggered in Phase 1 and Phase 2.

ios-clean