reproducing-ci-locally
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides documentation and guidance for a common developer task: synchronizing local development environments with remote CI runners to debug test failures.\n- [SAFE]: The shell commands included in the guide (sed, grep, pytest, uv, npx, gh) are standard development utilities used for their intended purposes of inspecting files, querying status, and running tests.\n- [SAFE]: The skill encourages security best practices by recommending the use of specific, pinned versions for tools (e.g., ruff@0.16.4, prettier@3.8.3) and explicit environment building to avoid side effects from the host system.\n- [INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface where an agent or user reads and executes commands derived from repository workflow files (.github/workflows/*.yml). This process inherently trusts the contents of the configuration files within the repo.\n
- Ingestion points: Workflow YAML files in the .github/workflows/ directory are parsed for shell commands.\n
- Boundary markers: None explicitly defined; the skill suggests following the workflow's command order and environment verbatim.\n
- Capability inventory: The skill recommends running various shell-based tools (pytest, ruff, npx, uv) based on the extracted workflow steps.\n
- Sanitization: None; the goal is the direct reproduction of the commands defined in the workflow contract.
Audit Metadata