The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from roadmap documents, external repositories, and web content, identifying them as potential attack surfaces. \n
Ingestion points: roadmap/<slug>/README.md, references/grilling-method.md, and content fetched from external registry/web sources via background investigators. \n
Boundary markers: The instructions explicitly mandate treating all read content as evidence/data rather than instructions and require the agent to flag any discovered embedded instructions. \n
Capability inventory: The skill possesses the ability to write to the roadmap directory, perform git commits and pushes, and execute local scripts via the shell. \n
Sanitization: The instructions explicitly forbid copying secret values into artifacts or logs, requiring citations instead, and mandate that answers be recorded faithfully to avoid agent-generated assumptions. \n- [COMMAND_EXECUTION]: The skill executes a local entrypoint scripts/brainstorm-state.ts using the bun runtime to manage roadmap states. \n
The agent is instructed to perform significant security validations before execution, including verifying absolute paths and using lstat to check for symbolic links or non-regular files to prevent unauthorized execution through path manipulation.

tailrocks-brainstorm