A Claude Code review workflow for pull-request changes, with source references and separate local feedback and optional publication steps.
The package documents a review command that examines changed code with project guidance and repository history. Its README describes parallel review perspectives and confidence filtering. Those are publisher-described mechanisms, not independently measured accuracy, and the AI Craft Pad Score does not reproduce the plugin's own confidence scale.
A maintainer who already has a reviewable pull request, clear repository rules and tests. The useful outcome is a small set of findings that a human can reproduce, rather than a claim that the change is correct.
Review quality depends on available context and the agent's permissions. Findings may be missed or mistaken. Source access and any optional comment-posting path need separate authorization; installing a review workflow does not authorize publication or merging.
Begin on a test pull request with one known defect. Ask for terminal-only feedback and inspect the cited code before requesting any external comment. Keep your ordinary test suite and reviewer approval in the process.
Check whether the tool identifies the seeded issue, points to the right revision and avoids treating old unrelated problems as new defects. Record a false positive as well as a success. This is a documentation-based assessment without a runtime benchmark.
claude-code
Publisher manifest for claude-code; runtime behavior not tested
Review the publisher setup, permissions, credentials and service costs before use.
Open the pinned package below and follow its publisher instructions for your agent. A Skills installer is not a universal plugin installer. Review hooks, bundled MCP connections, permissions and configuration before enabling the package.
plugins/code-review/.claude-plugin/plugin.jsonPresence is a static file/manifest observation. A missing declaration does not prove the component is absent at runtime.
Unknown; source software and connected services may have separate terms.
Unknown — no runtime security assessment.
Only sanitized manifest/registry metadata and bounded documentation are inspected. Package code, transitive dependencies, hooks and remote services have not been audited. Review access before use.
Category-based discovery suggestions. Review functional overlap; these combinations have not been integration-tested.
A structured investigation workflow for a failing build, unexpected application behavior or a regression that keeps returning.
Repository star change: +698 over 26.7 hours. 2026-11-01
A Python UI reference for Gradio interfaces, component layouts, event handlers and interactive model demonstrations.
Repository star change: +2 over 26.7 hours. 2026-11-01
A project scaffold and bundling workflow for complex React-based HTML artifacts.
Repository star change: +586 over 26.7 hours. 2026-11-01
Category-based discovery suggestions. Review functional overlap; these combinations have not been integration-tested.
A multi-format plugin bundle that organizes planning, testing and implementation through reusable development skills and startup guidance.
A Claude Code package for connecting Context7's hosted documentation lookup, useful when library versions matter to a coding task.
A Claude Code integration package that declares a GitHub MCP connection; review the bundle before adding a second GitHub server separately.
Category-based discovery suggestions. Review functional overlap; these combinations have not been integration-tested.
Explore freellmpool as an MCP connection for development. Compare the declared transport, package requirements and source evidence before connecting an agent.
Explore agents-party as an MCP connection for development. Compare the declared transport, package requirements and source evidence before connecting an agent.
Explore substack-ops as an MCP connection for development. Compare the declared transport, package requirements and source evidence before connecting an agent.
Choose a complementary capability only after checking for duplicated permissions and bundled connections. Agent Stacks explain the roles and validation steps; they are proposals, not tested integrations.
Build a stack for your task · Related AI guides · Report a correction