agentsclimarketplace

Dependency review

Skill olgaiv39/claude-oss-skills/skills/dependency-review

Practical Claude Code runbooks for clean, resource-efficient open-source development - from planning and implementation to debugging, review, release, and deployment

Install
npx -y skills add olgaiv39/claude-oss-skills --skill dependency-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 15 days oldThe repository was created 15 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 12 stars12 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Evaluate whether to add a new runtime or development dependency to a public open-source project before adding it. Use whenever a new package is proposed. Weighs standard-library alternatives, cost, transitive impact, maintenance, license, supply-chain risk, and low-resource effect, then decides add, do not add, or defer. Does not install anything.

SKILL.md

5.3 KB, as published. Nobody here has run it

dependency-review

Decide whether to add a proposed dependency before it is added. Do not install anything or modify project files while using this skill.

Activate when

  • A new runtime or development dependency is proposed
  • An existing dependency upgrade would pull new transitive packages
  • A failing build suggests adding a package to resolve it

Do not activate when

  • The package is already a project dependency and no version changes
  • The task is a code change with no new dependency -> use implement-minimal
  • The task is a failing test unrelated to dependencies -> use test-and-debug

Required inputs

  • The exact package name and, if known, the version or range
  • The problem the dependency is meant to solve
  • Whether it is a runtime or development dependency

Low-resource policy

Read the first of these that exists, then follow it:

  • ${CLAUDE_PROJECT_DIR}/.claude/shared/LOW_RESOURCE.md
  • $HOME/.claude/shared/LOW_RESOURCE.md

If neither exists, apply this fallback: run one expensive command at a time, prefer the narrowest validation, disable watch mode, reuse existing environments, and run full validation only at a milestone boundary. Do not scan the whole filesystem to locate the policy.

Context-efficiency policy

Read the first of these that exists, then follow it:

  • ${CLAUDE_PROJECT_DIR}/.claude/shared/CONTEXT_EFFICIENCY.md
  • $HOME/.claude/shared/CONTEXT_EFFICIENCY.md

If neither exists, apply this fallback: select files before reading; use targeted searches and bounded ranges; do not preload references; do not reread unchanged files; finish one atomic increment and stop; create a compact handoff before context is exhausted.

Facts that must not be assumed

  • The package manager and lockfile in use
  • That the package is actively maintained
  • That the license is compatible with public release
  • The size of the transitive dependency tree

Preflight

  1. Identify the ecosystem and manifest -> references/ecosystem-review.md
  2. Read the existing manifest to see current dependencies and conventions
  3. Determine whether a current dependency or platform API already solves it

Workflow

  1. State the exact problem the dependency would solve
  2. Check whether the standard library or platform API is sufficient
  3. Check whether an existing project dependency already solves it
  4. Estimate the local code avoided versus the code introduced
  5. Assess the package per ecosystem signals -> references/ecosystem-review.md
  6. Assess the low-resource impact -> references/low-resource-impact.md
  7. Confirm license compatibility for public release
  8. Weigh maintenance activity and supply-chain exposure
  9. Reach exactly one decision
  10. Produce the record using templates/dependency-decision.md
  11. Stop; installation happens later under implement-minimal if approved

Evaluation criteria

  • The exact problem being solved
  • Whether the standard library or platform API is sufficient
  • Whether an existing project dependency already solves it
  • Approximate amount of local code avoided
  • Runtime cost
  • Installation and build cost
  • Transitive dependency impact -> references/ecosystem-review.md
  • Maintenance activity and release recency
  • License compatibility for public release
  • Security and supply-chain exposure
  • Effect on low-resource development -> references/low-resource-impact.md
  • Effect on bundle or container size

Rules

  • Prefer platform APIs, existing dependencies, and small local functions
  • Do not reimplement cryptography, wallet security, authentication protocols, or mature parsers merely to avoid a dependency
  • Do not install the dependency as part of this review
  • Do not modify project files as part of this review
  • Do not run a command that downloads the package to inspect it

Decision branches

  • Standard library or an existing dependency suffices -> do not add
  • The package is sound, licensed, and maintained, and it avoids substantial correct-by-construction code -> add
  • Maintenance, license, or transitive impact is unclear -> defer pending evidence and name the evidence needed

Stop conditions

  • License compatibility cannot be confirmed
  • The package or its transitive tree cannot be inspected without installing
  • The decision requires human review (see below)

Human review boundaries

  • Any dependency touching cryptography, wallets, authentication, or key storage
  • A copyleft or unclear license on a package intended for public release
  • A package with an unknown or unverifiable maintainer or provenance

Final report

Produce the record in the exact section order of templates/dependency-decision.md, ending with exactly one decision: add, do not add, or defer pending evidence. Then stop.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.