The Agent Skills Directory

[COMMAND_EXECUTION]: Executes several internal binaries and scripts located in ~/.claude/skills/gstack/bin/ to manage project state, telemetry, and configuration settings.
[COMMAND_EXECUTION]: Starts a local HTTP server on a random port to serve a design comparison board, allowing the user to view and rate generated design variants in their browser.
[COMMAND_EXECUTION]: Uses eval and source to load environment variables and shell functions from the output of local scripts, a common pattern for integrated CLI tools.
[DATA_EXFILTRATION]: Contains logic for opt-in telemetry that logs skill usage data (skill name, duration, and repository basename) to local files and potentially an external endpoint.
[PROMPT_INJECTION]: Ingests local project documentation and prior session feedback to provide context for AI-driven design generation.
Ingestion points: Reads DESIGN.md, approved.json, and feedback.json from the filesystem.
Boundary markers: None; data from these files is interpolated directly into prompts for subagents without explicit delimiters or safety instructions.
Capability inventory: The skill is granted access to Bash, Agent, Read, and Write tools.
Sanitization: No sanitization or validation is performed on the content of ingested files before they are processed by the AI.

design-shotgun