The Agent Skills Directory

[SAFE]: The skill is a legitimate development utility focused on establishing repository standards and local safety environments. It includes detailed reference guides and templates for clean code, security boundaries, and automated testing.
[INDIRECT_PROMPT_INJECTION]: The skill reads external data from the local repository (e.g., existing documentation, package.json, and source files) to identify the stack and configuration. While this exposes a surface for indirect prompt injection if the ingested files contain malicious instructions, the risk is low due to the structured 'Explore -> Confirm -> Write' workflow and the human-in-the-loop requirement for confirmations.
Ingestion points: Reads package management lockfiles, framework configurations, and legacy documentation (e.g., DATABASE.md, BACKEND.md).
Boundary markers: Uses a specific Task Progress workflow to separate exploration from writing.
Capability inventory: Writes markdown documentation, generates linter configurations, and creates local utility scripts.
Sanitization: Relies on the agent following strict internal templates for output generation.
[DYNAMIC_EXECUTION]: The skill generates and writes local orchestration scripts (such as verify-touched.mjs) and configuration files based on predefined templates. This is an expected behavior for a bootstrapping skill intended to set up local development belts like linting and unit testing.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe practices by explicitly instructing the agent to never commit secrets or API keys and to use .gitignore for local configuration files.

setup-repo-standards