decaf-quality

Plugin decaf-quality from marketplace decaf in alphaleonis/decaf-claude. Commands read from the repo's manifest — 6 days ago.

Skills (11)

• auto-code-review

  Automated review-fix-recheck loop. Runs code review, triages findings, fixes autonomously via subagent, and re-reviews when the fixes warrant it. Iterates until code stabilizes.

• code-review

  Run parallel code review agents and consolidate findings into a unified report

• coherence-audit

  Audit a codebase for places where documentation, specs, comments, config, names, or versions disagree with the actual code, then resolve each (update docs, flag code, or accept). Use for documentation/spec audits, after major changes, or before a release.

• coverage-review

  Run code coverage analysis and review gaps for severity and test suggestions

• diagnose

  Investigate why something is broken and identify the root cause through competing hypotheses and evidence — diagnoses only, never proposes fixes. Use before attempting a fix, when the cause isn't obvious.

• refactor

  Analyze code for structural improvement opportunities and produce a prioritized refactoring plan

• resolve-code-review

  Walk through code review findings one at a time and decide a resolution for each — fix, skip, dismiss, or defer. Use "auto" for autonomous resolution with TDD.

• resolve-coverage-review

  Walk through coverage review gaps one at a time and decide a resolution for each — write tests, skip, dismiss, or defer. Use "auto" for autonomous test-writing.

• resolve-pr-feedback

  Walk through unresolved PR review feedback and decide a resolution for each thread — fix, reply, decline with evidence, or escalate. Works with Azure DevOps and GitHub. Use "auto" for autonomous resolution; replies are drafted and approved before posting.

• resolve-refactor

  Walk through refactoring plan opportunities one at a time and decide a resolution for each — apply, apply incrementally, skip, dismiss, or defer. Use "auto" for autonomous application.

• write-comments

  The rules for writing code comments and doc comments in any language, plus the project's glossary and local rules. Use before writing or editing source code, so new comments follow the rules from the start, or with a target to rewrite the comments already there.

Subagents (25)

• adversarial-reviewer

  Adversarial reviewer that constructs emergent failure scenarios — assumption violations, composition failures, cascade chains, abuse cases — in the space between pattern-matching reviewers. Dispatch — when the changeset has ≥50 changed executable lines, or touches a high-risk domain (auth, payments/

• broad-reviewer

  Broad code reviewer combining confidence scoring, knowledge preservation, and architectural analysis across 5 categories. Dispatch — always; part of the review floor in every mode.

• coherence-analyst

  Cross-file structural analyst for refactoring opportunities — duplication, naming consistency, validation scattering, interface drift, module boundaries, zombie code. Dispatch — spawned by the refactor skill (deep mode) to score cross-file improvements; not a code-review orchestrator persona.

• consistency-reviewer

  Sibling-consistency reviewer that compares the change against sibling code to catch unwritten-convention drift — naming/vocabulary, canonical-helper usage, attribute symmetry, leftover scratch code, comment-code mismatches, duplicated literals. Every finding must quote its convention source. Dispatc

• coverage-reviewer

  Assess which code-coverage gaps actually matter and suggest targeted tests. Dispatch — spawned by the coverage-review skill on parsed coverage gaps; not a code-review orchestrator persona.

• cpp-reviewer

  C/C++ stack reviewer for language-idiom misuse — lifetime and ownership errors, undefined behavior, RAII violations, exception safety, concurrency hazards. Dispatch — hard gate on file presence (C/C++ files in the changeset, never otherwise, under any preset), then a judgment gate on idiom surface (

• data-migration-reviewer

  Database migration reviewer for EF Core and SQL — schema drift, irreversible operations, data loss, missing backfills, deploy-window breakage, hot-table locking, rollback safety. Dispatch (hard gate) — only when migration artifacts are present in the diff (EF Core Migrations/*.cs, ModelSnapshot, .sq

• debugger

  Analyzes bugs through systematic evidence gathering - use for complex debugging

• design-reviewer

  System-level design reviewer for API contracts, data models, boundary violations, concurrency design, and evolution readiness. Evaluates design qualities observable in code. Dispatch — when public APIs, contracts, module boundaries, data models, or concurrency surface change; skip for single-functio

• dotnet-reviewer

  C#/.NET stack reviewer for language-idiom misuse — async/await pitfalls, disposal, EF Core change tracking, LINQ deferred execution, nullable reference types, threading. Dispatch — only under code-review's `audit` preset, or when named explicitly, and then only with C# files in the changeset (hard g

• finding-validator

  Independent validator for a single consolidated code-review finding. Adversarially re-verifies one finding against the code and returns confirmed/refuted/uncertain. Dispatch — one per surviving primary finding during the code-review validation wave; not part of the review roster.

• fix-verifier

  Cheap scoped check of one auto-code-review fix round. Verdicts each finding the round fixed as ADDRESSED or NOT ADDRESSED with file:line evidence, and inspects the round's delta for new breakage. Dispatch — only by auto-code-review Step 5.5, for a round that gets no full re-review but fixed a Medium

• go-reviewer

  Go stack reviewer for language-idiom misuse — goroutine leaks, error-handling discipline, typed-nil interfaces, channel misuse, context propagation, defer pitfalls, slice aliasing. Dispatch — hard gate on file presence (Go files in the changeset, never otherwise, under any preset), then a judgment g

• knowledge-reviewer

  Reviews code for knowledge preservation — undocumented decisions, implicit assumptions, comprehension risks, and lost context that makes code unmaintainable. Dispatch — any substantive change; skip only purely mechanical diffs (formatting, renames, typo fixes, generated files).

• performance-reviewer

  Performance reviewer for throughput, latency, and resource cost — N+1 queries, unbounded memory growth, hot-path work, missing pagination, algorithmic complexity on unbounded input. Dispatch — when the diff contains database/ORM queries, loops over collections with I/O or allocation, async/concurren

• pr-thread-resolver

  Resolves a single PR review thread — verifies the feedback against current code, implements or declines with evidence, and returns a verdict with a drafted reply. Dispatch — one per actionable thread during resolve-pr-feedback's auto mode; not part of the review roster.

• prior-feedback-reviewer

  Reviews a PR diff against the PR's existing review threads — unaddressed feedback, partially addressed requests, and regressions of prior fixes. Dispatch (hard gate) — only when reviewing a PR that has prior human review threads; never spawned otherwise, under any preset.

• quick-reviewer

  Fast generalist code reviewer for bugs, logic errors, security patterns, code quality, and project convention violations. Complements deeper specialized reviewers. Dispatch — only under code-review's `audit` preset, or when named explicitly; a second generalist next to broad-reviewer, which owns the

• rust-reviewer

  Rust stack reviewer for language-idiom misuse — panic paths in production code, unsafe-block invariants, async hazards (blocking calls, guards across await), lock discipline, error-context erasure. Dispatch — hard gate on file presence (Rust files in the changeset, never otherwise, under any preset)

• security-reviewer

  System-level security reviewer for threat modeling, architectural security gaps, and missing controls. Complements broad-reviewer's pattern-level security checks. Dispatch — when the diff contains a concrete trust-boundary trigger — a handler/route/middleware, parsing of data crossing a process/user

• solo-reviewer

  Whole-surface single-pass reviewer — the sole seat of the `bugs` preset. Reads the entire changeset across every concern class, verifies findings empirically inline, self-calibrates severity and confidence, and returns a complete report carrying full responsibility for the verdict. Dispatch (hard ga

• spec-compliance-reviewer

  Spec compliance reviewer that verifies implementation matches a previously drafted specification or plan. Maps requirements to code, identifies gaps, deviations, and scope creep. Dispatch (hard gate) — only when a spec is available, provided via --spec or discovered by the skill (PR-linked work item

• structural-analyst

  Per-file structural quality analyst for refactoring opportunities — naming, composition, complexity, domain modeling, type design, error handling. Dispatch — spawned by the refactor skill to score per-file structural improvements; not a code-review orchestrator persona.

• test-reviewer

  Expert test code reviewer. Use PROACTIVELY after writing or modifying tests. Reviews ONLY test files for silent failures, false positives, flaky patterns, and anti-patterns. Does NOT review production code. Dispatch (hard gate) — only when test files are present in the changeset; never spawned other

• typescript-reviewer

  TypeScript/JavaScript stack reviewer for language-idiom misuse — floating promises, type-system escape hatches, equality coercion, runtime boundary trust, event-loop blocking, shared-state mutation. Dispatch — hard gate on file presence (TypeScript/JavaScript files in the changeset, never otherwise,

Hooks (2)

• SessionStart

  node ${CLAUDE_PLUGIN_ROOT}/hooks/comments-pointer.js

• SubagentStart

  node ${CLAUDE_PLUGIN_ROOT}/hooks/comments-pointer.js

Similar plugins