The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands, including find, grep, and git log, to analyze codebase structure, file contents, and history. These commands are used for static discovery and diagnostic purposes (e.g., finding project manifests or architectural leaks) and are scoped to the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret data from a potentially untrusted codebase (source code, documentation, and manifest files), which creates a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: The agent reads project manifests (package.json, csproj, etc.), architectural documentation (CLAUDE.md, README.md), and source code across the directory tree using shell commands.
Boundary markers: The skill architecture mitigates risk through a strict human-in-the-loop design, incorporating six explicit "Stop-and-ask" checkpoints that require user confirmation before finalizing project lists, classifications, or writing the blueprint artifacts.
Capability inventory: The skill possesses the ability to read files, discover directory structures, and write documentation files (BLUEPRINT.md, SYSTEM_MAP.md). It does not have network access or administrative privileges.
Sanitization: The skill instructions emphasize "Discovery over theory" and "Evidence over assertions," requiring the agent to cite specific file paths and line numbers for all claims, which facilitates manual verification by the user.

project-blueprint