defect-review · v0.16.0 · 2026-11-15 · sha256 5329d467be2acb55
defect-review v0.16.0A
Immutable. This exact content never changes and is served at /api/v1/blob/5329d467be2acb55.
No tool to install first
claude plugin marketplace add https://markdownregistry.com/mp/a/art_zkvgwilbbfbo4e2m/5329d467be2acb55e45c0f389d308c8a947c8bb0a5d2fa99064f78b4560311c9.json
Then /plugin install defect-review@markdownregistry-zkvgwilbbfbo4e2m in Claude Code. It installs this exact version, the one graded A here, pinned to the commit it was graded at, and leaves it there.
--- name: defect-review description: "Explicit Defect Review of Go, TypeScript, Rust, or Java PRs or local source to find bugs by reading, or help using that review." disable-model-invocation: true metadata: version: "0.16.0" --- # Defect Review **Find bugs.** Find places where the selected source does the wrong thing for someone: a caller gets a wrong result, a failure is reported as something it is not, a wait outlives its limit, something started is never ended, one caller's trouble damages another's work. A finding traces the behavior through the source, line by quoted line, says what should have happened and on what basis, and has survived an independent attempt to show it is not a bug. A written promise is one basis among several, not a precondition. A document that describes harmful behavior does not make it right. The review reads. It does not clone, copy, build or run the project, fix what it finds or post comments. Readability, design taste, security and test quality are other reviews' subjects. ## Run For help, explain usage and stop. If no target can be identified from the request, ask only for that. Otherwise orchestrate through the active adapter, [Claude Code](references/harnesses/claude.md), [Codex](references/harnesses/codex.md), [Antigravity](references/harnesses/antigravity.md) or [OpenCode](references/harnesses/opencode.md), following [the workflow](references/workflow.md). When the user explicitly asks to run agents in another installed harness, such as OpenCode, start them there through [external agents](references/harnesses/external.md). Each reader gets one lens, or a related pair for a small change, and [the contract](references/contract.md). Each challenger gets one reader's claims with their proofs and [the challenge](references/challenge.md), never the reader's reasoning. You decide every candidate yourself and deliver [the report](assets/report-template.md) in the user's language. If agents cannot be started, say the review is a limited single-reader pass and do not claim an independent challenge.