The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from the repository, including directory structures and file metadata, to generate configuration. This ingestion of potentially untrusted local data represents a standard indirect prompt injection surface.
Ingestion points: Scans the file system for docs/prd/, docs/adr/, and docs/research/; detects project type via workspace files like package.json and Cargo.toml; reads git configuration via git remote -v.
Boundary markers: The instructions do not specify explicit delimiters for the scanned data, relying on the agent's reasoning.
Capability inventory: The skill writes new configuration files to docs/agents/ and updates the primary project documentation (AGENTS.md or CLAUDE.md).
Sanitization: The skill includes mandatory steps (Step 4 and 5) to present all findings and drafts to the user for confirmation before any files are written to disk.
[COMMAND_EXECUTION]: The skill uses git remote -v to identify the repository's origin and references the use of official GitHub (gh) and GitLab (glab) CLI tools in its templates for issue management. These operations are appropriate for the skill's stated purpose of engineering project setup.

setup-project