node-deps-bumper
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local helper scripts (run-taze.sh, update-bun-catalogs.py) and various development tools (taze, ni, and the target project's own build and test suites) to perform and verify dependency updates.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including package.json files, dependency update plans, and external migration/release notes. This represents a potential attack surface where malicious instructions could be embedded in package metadata to influence the agent's behavior during the update process.
- Ingestion points: The skill reads manifest files and instructs the agent to review external release/migration notes for major updates.
- Boundary markers: The instructions define structured table formats for presenting plans but do not specify explicit delimiters or "ignore" instructions for the untrusted text being processed.
- Capability inventory: The agent has the capability to execute shell commands, run scripts, and write to the filesystem.
- Sanitization: No specific sanitization or filtering is defined for the external metadata or release notes ingested by the skill.
Audit Metadata