The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the taze CLI tool to analyze dependencies and run updates. It also identifies and runs a repository's existing validation suite (e.g., tests, builds, lints) to ensure that updates do not introduce regressions.
[EXTERNAL_DOWNLOADS]: The skill references and utilizes the taze utility from the antfu-collective GitHub organization, a well-known source in the Node.js ecosystem. It provides instructions for installing the tool if it is missing from the environment.
[REMOTE_CODE_EXECUTION]: The skill executes local Python and shell scripts provided within its package using bash and uv run. These scripts are used for parsing plan data and managing Bun catalog updates in package.json.
[PROMPT_INJECTION]: The skill processes data from external package registries (via taze) to generate update plans. The risk of malicious input is mitigated by the use of strict regex validation for version strings and a mandatory human review step for all major version updates. Evidence Chain: 1. Ingestion points: Package names/versions from registries via taze output. 2. Boundary markers: Plan presented in structured markdown tables. 3. Capability inventory: Local file writes to manifests and execution of repository test/build scripts. 4. Sanitization: Use of VERSION_RE regex to validate all ingested version strings.

node-deps-bumper