ln-42-dependency-upgrader
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes system-native tools including package managers (e.g., npm, pip, yarn) and version control systems (git) to inventory, update, and verify dependencies. These actions are within the expected functional scope of a dependency management tool.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources such as package registries, manifests, and release notes. It addresses potential risks by instructing the agent to verify claims against official documentation and migration guides, and by requiring manual inspection of install scripts from new or high-risk dependencies.
- Ingestion points: Dependency manifests, lockfiles, and external registry metadata referenced in the 'Discover Scope' and 'Research and Plan Batches' sections.
- Boundary markers: The instructions explicitly mandate checking official release notes and migration guides rather than relying solely on registry defaults, creating a logical verification boundary.
- Capability inventory: The skill involves file read/write operations (manifests and lockfiles) and shell command execution (package managers, git, and project test suites).
- Sanitization: Verification is performed via repository-defined test suites and mandatory comparison with authoritative upstream documentation.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to run repository-defined scripts for building, testing, and linting to ensure project stability after upgrades. It includes a specific warning against running lifecycle scripts from untrusted package sources without appropriate environmental safeguards.
Audit Metadata