brent-tools

Plugin brent-tools from marketplace brentrector-claude-skills in BrentRector/claude-skills. Commands read from the repo's manifest — 26 days ago.

Skills (15)

• agent-fleet

  Use before dispatching more than a couple of parallel subagents or a multi-agent workflow on a long-running engineering campaign — so a session-limit kill costs at most one step, a restart never repeats work, and the token budget is spent where it moves the needle.

• architecture-audit

  Use when a whole codebase (not one diff) needs a comprehensive architectural review AND restructuring - god classes to split, duplicate code and duplicate rules to unify, file-system layout, namespaces, file and class names to fix, dependency direction to enforce, and the language and runtime to mod

• automating-agent-guardrails

  Use when a project's rules for Claude agents keep getting broken or skipped (forbidden git commands, direct pushes to a protected branch, read-only reviewers that write, owner-approval steps nobody asks about), or when setting up or auditing Claude Code guard hooks, `.claude/agents` role definitions

• claude-cloud-sessions

  Use when running work in Claude Code cloud sessions (claude.ai/code, `claude --cloud`, routines/RemoteTrigger) — setting up an environment, launching and billing sessions, multi-repo and private-submodule access, orchestrating subagents in the cloud, and stopping cleanly before a credit or budget ru

• cross-platform

  Use when code, tests, scripts or agents must work on both Windows and Linux (including WSL), when CI runs on an operating system the local gates do not, when a Windows-hosted agent edits files or runs shell commands, or when something passes on one OS and fails on the other. Covers running the targe

• devlog

  Use when a project keeps (or should keep) a development log, work history, changelog of findings or fix log that Claude agents write and read, including setting one up, writing, numbering or timestamping an entry, recording a failed, reverted or overturned approach, checking whether something was tr

• dotnet-engineering

  Use for any .NET or C# work - writing or reviewing C#, running or filtering dotnet test, writing tests or judging whether tests would catch a bug, BenchmarkDotNet measurements, diagnosing an MSBuild failure from a binlog, trimming or Native AOT warnings (IL2026, IL3050 and friends), or publishing a

• engineering-standards

  Use whenever writing, changing, designing or reviewing production code - it states the non-negotiable quality bar (commercial-grade, decades-maintainable), the structural rules (no god classes, one mechanism per job, one rule in one place), the root-cause and every-bug-is-a-pattern discipline, and w

• performance-diagnosis

  Use when something is slower than it should be - a test suite, a build, a compile, a service, a batch job - when adding threads or machines stops helping, when CPU sits idle while work waits, or before any optimization at all. Measure first, then find the mechanism; a scaling curve, a processes-vers

• review

  Use when asked for a code review, architecture review, performance review, or duplication/efficiency analysis of a diff, branch or PR (or a whole subsystem audit), including the read-only review of a finished branch before it is enqueued to land - runs the four review dimensions (architecture, full

• roslyn-analysis

  Use in C#/.NET code when a review or refactor needs facts from the compiler rather than from grep - finding duplicated method bodies and near-clones across a solution, listing every implementation, override, caller and reference of a mechanism for a sibling sweep, applying one mechanical change acro

• spec-compliance-audit

  Use when auditing a whole implementation against a written standard (language standards, ECMA-335, RFCs, file formats, protocol specs) - decompose the standard into a rule catalog, one agent per rule with a self-contained input, a fixed verdict vocabulary, mechanically checked citations, an adversar

• spec-oracle

  Use BEFORE implementing, debugging or deciding any behavior governed by a written standard or spec (language standards, protocols, file formats, RFCs, ECMA-335, ...) — derive the expected behavior from the spec and produce a checked citation before reading or writing code.

• test-gate

  Use before every commit, merge, enqueue or push to choose and run the right test gate — a fast targeted local gate for every change, with CI's merge-group run as the whole-population regression net — and to read the result without producing a false green.

• variant-analysis

  Use immediately after any defect is confirmed - a wrong answer, crash, security hole, or bad pattern - and whenever someone asks "are there others like this?" or "is this the same bug?". Names the mechanism rather than the symptom, derives textual, structural (Roslyn, ANTLR, tree-sitter, Semgrep) an

Subagents (5)

• README
• comment-analyzer

  Use when a change adds or modifies comments, docstrings, design-doc references or citations of a spec, RFC, issue or standard - verifies each claim against the code and each citation against its source, and flags comments that lie, justify an omission with an unrelated reference, or have gone stale.

• pr-test-analyzer

  Use when a change adds or modifies tests, or adds behavior that tests should cover - checks that the tests verify the COMPLETE behavior the change claims (not just the one bug that prompted it), that expected values come from the governing authority rather than copied output, that the new tests are

• silent-failure-hunter

  Use when a change touches error handling, catch blocks, fallbacks, default values, retries, optional chaining/null-coalescing, feature flags, "not implemented" stubs, filters or gates - or on any review where a wrong answer would be worse than a crash. Hunts for code that produces a plausible result

• type-design-analyzer

  Use when a change introduces or reshapes types, classes, records, enums, interfaces or module boundaries - flags god classes, stringly-typed state, primitive obsession, duplicated mechanisms, invariants enforced only by convention, and illegal states that the type makes representable. Reports only f