The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Bun runtime environment by piping a shell script from the official bun.sh domain directly to bash if the runtime is not detected locally.
[EXTERNAL_DOWNLOADS]: Fetches and executes the Bun installation script from the well-known bun.sh service. It also offers to open educational resources and application links for Y Combinator in the user's browser.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it reads and processes external data from the project repository (such as CLAUDE.md, TODOS.md, and DESIGN.md) and git history to inform its design recommendations.
Ingestion points: Reads project documentation (CLAUDE.md, TODOS.md, DESIGN.md), existing design documents in ~/.gstack/projects/, and git log/diff outputs.
Boundary markers: No specific boundary markers or 'ignore' instructions are used when interpolating file content into the agent's context.
Capability inventory: The skill has access to tools for command execution (Bash), file modification (Write, Edit), and web searching.
Sanitization: There is no evidence of sanitization or validation of the content read from external files before it is used to influence the skill's logic.
[COMMAND_EXECUTION]: The skill executes multiple local binaries and scripts from the ~/.claude/skills/gstack/bin/ directory for configuration, update checks, and telemetry logging. It also executes a local ./setup script during initial configuration.
[DATA_EXFILTRATION]: Contains a telemetry module that logs usage statistics (skill name, duration, session ID, and success outcome) to a local directory and periodically transmits this data to a remote service. The skill provides clear notice to the user about this behavior and includes an opt-out mechanism.

office-hours