swe-workbench

C# and .NET idioms — .NET 8 LTS, csproj, C# NRT, records, value semantics, switch expressions, async/await, Task, ValueTask, CancellationToken, ConfigureAwait, dependency injection, IOptions<T>, LINQ, Span<T>, ArrayPool, hot path, allocation profiling, and performance. Auto-load when working with .cs files, .csproj, .sln, Directory.Build.props, or when the user mentions C#, dotnet, .NET 8, C# NRT, NRT, records, value semantics, switch expressions, Task, ValueTask, CancellationToken, ConfigureAwait, IOptions, LINQ, Span, or hot-path code.

Dart idioms — null safety, async/await, Futures and Streams, widget composition, and Riverpod/Bloc app architecture. Auto-load when working with .dart files, pubspec.yaml, or when the user mentions Dart, Flutter, widgets, Riverpod, Bloc, Streams, or null safety.

Go idioms — error handling, concurrency, and standard library usage. Auto-load when working with .go files, go.mod, go.sum, or when the user mentions Go, Golang, goroutines, channels, interfaces, context, or error wrapping.

Java idioms — records, sealed types, virtual threads, streams, and JDK 21+ patterns. Auto-load when working with .java files, pom.xml, build.gradle, or when the user mentions Java, JVM, Spring, Maven, Gradle, records, sealed classes, or virtual threads.

Kotlin idioms — null safety, coroutines, sealed interfaces, scope functions, and Flow. Auto-load when working with .kt files, build.gradle.kts, or when the user mentions Kotlin, coroutines, suspend, StateFlow, sealed interface, or Kotlin DSL.

Python idioms — PEP 8, type hints, dataclasses, context managers, generators, asyncio, and testing. Auto-load when working with .py files, pyproject.toml, requirements.txt, or when the user mentions Python, pytest, asyncio, dataclass, type hints, or virtualenv.

Ruby idioms — Ruby 3.x pattern matching, blocks, procs, lambdas, Enumerable, Comparable, frozen string literals, error handling, and testing. Auto-load when working with .rb files, Gemfile, Rakefile, gemspecs, or when the user mentions Ruby, Bundler, RSpec, minitest, blocks, procs, lambdas, or pattern matching.

Rust idioms — ownership, borrowing, lifetimes, error handling, traits, and iterators. Auto-load when working with .rs files, Cargo.toml, or when the user mentions Rust, cargo, ownership, borrow checker, trait, lifetime, or async Rust.

Swift idioms — optionals, value types, actors, async/await, protocols, and Result builders. Auto-load when working with .swift files, Package.swift, or when the user mentions Swift, SwiftUI, actors, async/await, Sendable, Result builders, or Swift Package Manager.

TypeScript and JavaScript idioms — strict mode, type safety, discriminated unions, async patterns, and Node. Auto-load when working with .ts, .tsx, .js, .jsx files, package.json, or when the user mentions TypeScript, JavaScript, Node, type safety, or tsconfig.

Clean code, DRY, KISS, YAGNI, function length, naming, abstraction level, error handling, function argument count, Command-Query Separation, Boy Scout Rule, intent-revealing naming. Auto-load when writing functions, naming variables, reviewing code clarity, discussing comments, or debating whether to abstract.

Code review heuristics — five-axis review lens (correctness, security, design integrity, test coverage, comment quality); confidence-based filtering (no finding without a concrete failure scenario); review comment tone (observation over accusation); nitpick filtering; what counts as a real finding vs linter noise. Auto-load when writing or framing a review comment, deciding whether a PR finding is worth surfacing, reviewing a diff for correctness, or filtering review nitpicks.

Development workflow — full lifecycle from Branch → Implement → Verify → Review → Deliver. Activated by /swe-workbench:implement, /swe-workbench:design, /swe-workbench:refactor, /swe-workbench:debug, /swe-workbench:test, /swe-workbench:architect, /swe-workbench:migrate, and /swe-workbench:document when the plan being authored modifies the codebase (Mode A) or when driving an implementation (Mode B). Entry point to execute a written implementation plan end to end — delegates to superpowers:executing-plans and superpowers:subagent-driven-development with the full 5-phase lifecycle. Skip for pure design / analysis output. Can also be invoked directly to author a Workflow section, run the 5-phase implementation flow, or orchestrate parallel agents (Mode C).

Use after a PR has been merged on GitHub to remove the local worktree, delete the local branch, delete the remote branch, and fast-forward local main — safely, with squash-merge support.

Grill-me interrogation mode: relentlessly walk the decision tree one question at a time, interrogate requirements, self-answer from the codebase, exit on shared understanding or proceed — hands a Resolved-decisions block to the command's artifact step. Activated by /swe-workbench:capture /swe-workbench:design /swe-workbench:implement /swe-workbench:architect /swe-workbench:extend /swe-workbench:debug /swe-workbench:hotfix. Not a from-scratch design flow; produces no design doc.

Branch-based P0 hotfix lifecycle — skips the worktree provider, ships a PR immediately after a minimal fix (ahead of verification and review, deferred by design), stamps a deferred-verification marker on the PR body, then reconciles it once a regression test lands and the suite is green again. Activated by /swe-workbench:hotfix.

Fetch ticket context from Jira (PROJ-123), Linear, Confluence, or GitHub (`#NNN`). Auto-loads on ticket references. Returns title, summary, acceptance criteria, links, and recent activity.

Posting core shared by workflow-pr-review, workflow-pr-review-followup, and the specialist /swe-workbench:review PR-mode sub-flow — takes a normalized findings/decision/byline payload, dedupes against existing review threads (±5-line fuzzy match + Jaccard ≥ 0.4), posts new inline or PR-level comments, applies the self-review + diff-scoping decision flip, and submits APPROVE or COMMENT.

Use when the user mentions a worktree — to start, switch, resume, continue, or end a worktree session. Routes to `EnterWorktree` / `ExitWorktree` so the running claude moves into the worktree without restart. Covers switching into an existing worktree (`git worktree add` then mid-session entry), resuming or continuing work in an already-created worktree, pivoting current work into a new worktree, and exiting back. Also fires when a user describes the cd-prefix symptom ("I've been cd-ing into the worktree", "I've been cd-prefixing commands"). Does NOT replace `superpowers:using-git-worktrees` for new feature kickoffs (consent, ignore-check, baseline tests).

Use when a reviewer wants to re-check a PR after the owner has addressed feedback — re-runs the reviewer agent against the updated diff, deduplicates against existing threads (Jaccard ±5-line), posts only truly-new inline comments, and submits an APPROVE or COMMENT review event.

Use when reviewing a remote GitHub PR — fetches into an ephemeral worktree, runs the reviewer agent with a Review Decision footer instruction, deduplicates findings against existing review threads (±5-line fuzzy match + Jaccard ≥ 0.4 against any author), posts new inline comments via gh-api, adds 👍 reactions on dedup matches, and submits the review with APPROVE or COMMENT. Counterpart of local-diff review (which workflow-development Phase 4 keeps using).

Fetch on-call and discussion context from Slack threads and PagerDuty incidents — participants, timeline, and status. Auto-loads on Slack permalinks or PagerDuty incident references. Keywords: Slack, thread, permalink, PagerDuty, incident, on-call, participants.

Fetch production observability signals from Sentry — error, exception, stack trace, regression, and alert data. Auto-loads on Sentry references. Returns message/culprit, first-seen, frequency, and a condensed stack trace, always labeled as Phase-1 framing, never a profile.

Profile-first performance investigation runbook — reproduce and baseline a metric, pick a profiler from the ecosystem tooling matrix (Go/Python/JVM/Node/Rust/Ruby/.NET/native), capture CPU + alloc profiles under load, hand the profile to the performance-tuner agent for ranked hotspot read, classify by bottleneck taxonomy (CPU/memory-GC/I-O/lock-contention), form one hypothesis, make one isolated change, before/after measurement, regression guard. Keywords: profile flame graph hotspot bottleneck slow endpoint investigation benchmark.

Sync the current branch with the default branch via rimba or git, then resolve conflicts file by file with a per-hunk rationale and a keep-mine / keep-main / manual recommendation before staging. Push only happens after explicit confirmation — never automatic.

Use when the user wants to commit staged changes or create a PR — enforces trigger-phrase discipline (preview vs commit vs ship), the [type] commit format, branch-naming check, [no ci] auto-appended on docs-only commits, draft-vs-ready PR prompt, and PR template detection. Pre-merge counterpart to workflow-cleanup-merged.

SQL idioms — query tuning, EXPLAIN plans, table definitions, constraints, indexes, transactions, window functions, CTEs, and pagination. Auto-load when working with .sql files or migrations, or when the user mentions SQL syntax, SELECT, JOIN, EXPLAIN, CTE, window functions, transaction isolation, deadlocks, or SQL indexes.

Use when a PR owner wants to address review feedback — fetches outstanding threads and general comments, presents a per-item triage (ADDRESSED / CLARIFIED / DEFERRED), applies fixes via the Edit tool, commits via workflow-commit-and-pr, posts replies via the GitHub API, resolves addressed threads via GraphQL resolveReviewThread (those without a thread are never resolved), and syncs stale PR metadata when drift is detected.

Use when investigating a bug to root-cause and FILE a GitHub issue rather than fixing immediately — counterpart to /swe-workbench:debug (which patches). Enforces the Iron Law (no patches without root cause), runs a 4-phase loop (Investigation, Pattern Analysis, Hypothesis, File issue), and produces a structured issue with code-path table and impact assessment.

Use to understand an unfamiliar codebase — presents architecture overview, module map, public API surfaces, and conventions as a layered mental model. Knowledge-presentation only: read-only sweep, no defect ranking, no reasoning chains. Distinct from workflow-codebase-audit (finding-oriented defect detection) and tech-writer / /swe-workbench:document (generates new prose artifacts). Ideal for onboarding, knowledge-handoff, and reviewer context-gathering.

Captures a mid-PR sub-idea and implements it on the same branch as the existing PR — skips Phase 1 (Branch), preserves Verify → Review → Deliver, and uses workflow-commit-and-pr to update existing PR. Never creates a new branch or new PR.

Bash idioms — strict mode, quoting, parameter expansion, arrays, pipefail, trap cleanup, idempotency, heredocs, and POSIX portability. Auto-load when working with .sh, .bash files, or when the user mentions bash, shell, sh, shellcheck, set -e, or pipefail.

Domain-Driven Design (DDD) — bounded contexts, aggregates, entities, value objects, ubiquitous language, domain events, context map, anti-corruption layer, and repository pattern. Auto-load when modeling a complex domain, splitting services, deciding service boundaries, microservice splits, designing aggregates, or aligning code with business language.

Use when walking the dependency-upgrade lifecycle — routine sweep, Dependabot/Renovate PR triage, CVE patch, or major-version migration. Structured runbook: triage+batch → bump regen lockfile → build/test → breakage-triage → PR hygiene. Per-ecosystem command matrix; composes principle-security. Keywords: dependency upgrade bump lockfile Dependabot Renovate semver CVE supply-chain breakage

Performance engineering principles — latency vs throughput, profile-before-optimize discipline, Big-O instincts for common patterns, allocation and GC pressure, data locality and cache-friendliness, N+1 queries on a list endpoint, cache-aside and write-through invalidation, TTL sizing for hot-path endpoints, stampede prevention with single-flight. Auto-load when reviewing hot-path code, choosing data structures, designing batch or streaming pipelines, hunting allocations or GC pauses, weighing latency trade-offs, sizing TTL and picking cache invalidation, preventing cache stampede or thundering herd with single-flight on an expired hot key, detecting N+1 queries on list endpoint calls, or evaluating scalability.

Resiliency principles — fault tolerance, resilience, partial failure, blast radius, failure domains, bulkheads, resource isolation, graceful degradation, fail-fast vs fail-soft, health checks, liveness vs readiness probes, cascading failure, gray failure, fault isolation, idempotency, idempotency keys, safe retry, deduplication, exactly-once, rate limiting, throttling, quotas, backpressure, token bucket, leaky bucket, jitter, thundering herd, synchronized retries, 429. Auto-load when designing for partial failure, isolating dependencies via bulkheads, planning graceful degradation, choosing fail-fast vs fail-soft, configuring health/readiness/liveness probes, evaluating cascading failure risk, designing fallback paths, reviewing system-level fault tolerance, making operations safely retryable with idempotency keys, implementing dedup stores, designing rate limiters or token buckets, handling 429 responses and backpressure, adding jitter to avoid thundering herd retry storms and synchronized retry spikes, or p

Error handling principles — errors as values vs exceptions, error classification (transient/permanent/programmer), wrapping and context chains, sentinel vs typed vs coded errors, retry with exponential backoff and jitter, timeouts and deadlines, circuit breakers and bulkheads, log-once boundary discipline. Auto-load when discussing errors as values, Result type, exception handling, retry, exponential backoff, jitter, circuit breaker, fail fast, fail soft, idempotent retry, error wrapping, timeouts, or deadlines.

Design patterns — Gang of Four and beyond. Strategy, factory method, observer, decorator, adapter, facade, template method, command, repository, dependency injection. Auto-load when designing class structure, refactoring toward a pattern, or evaluating whether a pattern applies.

After a codebase audit, groups audit findings by subsystem (path prefix) and emits them as context-rich GitHub issues with template discovery, label selection, and a batch preview-confirm gate before filing. Keywords: audit findings grouped github issues subsystem emit.

UX and product design principles — Nielsen's 10 usability heuristics, visual hierarchy, information architecture, interaction design patterns, design-system compliance, and responsive layout. Accessibility/WCAG conformance is delegated to `swe-workbench:principle-accessibility`. Auto-load when reviewing UX, evaluating usability, checking interaction patterns, assessing visual hierarchy or information architecture, auditing design-system compliance, or planning user-facing features.