Skip to content

plugin-packages.md

Latest commit

 

History

History
96 lines (73 loc) · 2.93 KB

plugin-packages.md

File metadata and controls

96 lines (73 loc) · 2.93 KB
Raw

Plugin: plugin-packages

Local package manifest and lockfile parsing, evidence-backed package review, and workflow ownership.

Status:

Package: plugins/plugin-packages
NPM name: @security-workbench/plugin-packages
Quality: beta
Execution: local-only
Network: none
Persistence: none
External binaries: none
Implemented skills: 3
Owned workflows: 2

Purpose

plugin-packages owns the package-management domain boundary for local package.json and lockfile inspection. It parses observed metadata and emits evidence-backed inventory and install-surface signals without installing packages, executing scripts, contacting registries, or performing vulnerability lookup.

Implemented skills

Skill Input Output summary
parse_package_json package.json text package metadata, scripts, dependency sections, engines, repository, bin, and workspace presence
parse_lockfiles npm, pnpm, or Yarn lockfile text package inventory, versions, dependency edges, importers, and root dependency sections
review_package either parser output or its JSON run result evidence-backed package inventory and install-surface signals

Owned workflows

package_manifest_review
  parse_package_json
  → review_package

lockfile_review
  parse_lockfiles
  → review_package

Dependencies

@security-workbench/core-parsers
@security-workbench/schemas

The direct core-parsers dependency provides generic bounded text and native-JSON parser helpers. Package-domain interpretation remains inside this plugin.

Security boundary

no package installation
no lifecycle-script execution
no registry access
no package tarball retrieval
no vulnerability or reputation lookup
no reachability or exploitability claims
no risk scoring
no finding generation

Inputs and outputs may contain attacker-controlled package names, versions, paths, registry URLs, and script names. Treat all artifact-derived values as untrusted presentation data.

Examples

security-workbench workflows run package_manifest_review \
  --input-file "$PWD/fixtures/package-json/basic-package.json" \
  --format pretty

security-workbench workflows run lockfile_review \
  --input-file "$PWD/fixtures/lockfiles/package-lock.json" \
  --format pretty

Manual chains remain available:

security-workbench skills run parse_package_json \
  --input-file "$PWD/fixtures/package-json/basic-package.json" \
  > /tmp/package.parsed.json
security-workbench skills run review_package \
  --input-file /tmp/package.parsed.json \
  --format pretty

Limitations

  • Review is based only on observed manifest or lockfile metadata.
  • Script presence is not proof that a script is malicious or will execute.
  • Package names and versions are not validated against a registry.
  • Vulnerability, provenance, maintainer trust, and package reputation are not assessed.
  • Dependency reachability and exploitability are not determined.