The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected in the skill instructions or provided templates. The skill proactively includes security hardening advice such as pinning actions to SHAs and using minimal GITHUB_TOKEN permissions.
[EXTERNAL_DOWNLOADS]: The skill configures workflows that utilize official and well-known GitHub Actions from trusted organizations, including GitHub (actions/checkout, actions/setup-go), Astral (setup-uv), and Oven (setup-bun). These references are documented neutrally as they target verified services.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests untrusted project metadata to generate configurations.
Ingestion points: The skill scans project files such as pyproject.toml, package.json, Cargo.toml, and go.mod in the user's repository (SKILL.md, Step 1).
Boundary markers: None explicitly mentioned during the detection phase.
Capability inventory: The skill writes new YAML workflow files to the .github/workflows/ directory (SKILL.md, Step 3).
Sanitization: The skill utilizes static templates for generation, which limits the influence that malicious metadata could have on the resulting CI configuration.
[COMMAND_EXECUTION]: The skill provides templates and pre-commit hooks that execute standard development tools (e.g., ruff, pytest, biome, bun test) via shell commands. These are legitimate operations for a CI/CD scaffolding tool.

ci-scaffolding