superpowers-extensions

Use when the user asks to compress a natural-language memory file (CLAUDE.md, STATE.md, ROADMAP.md, project notes) to save input tokens. Preserves code blocks, URLs, file paths, frontmatter, headings, tables, and list structure byte-exact. Backs up the original to FILE.original.md before each compression. Refuses to operate on downstream-consumed artifacts (plan documents, UI contracts, impact analyses, design documents, review reports). Triggers: "/compress-memory <path>", "compress STATE.md", "shrink CLAUDE.md", "compact this memory file". Also auto-invoked by project-orchestration:pause-work when ROADMAP.md frontmatter contains compress_memory: enabled.

Use for multi-session project lifecycle management. Triggers on kickoff phrases ("plan the roadmap", "new project", "start a project from scratch", "greenfield", "what milestones do we need"); session lifecycle ("resume", "continue", "where are we", "pause", "I'm done for today"); phase/milestone ops ("add a phase", "new milestone", "audit the milestone", "complete the milestone"); session start on a project with a docs/planning/ directory; or an empty/greenfield repo where the user signals multi-milestone work. Covers greenfield kickoff via plan-roadmap (whole-project brainstorm covering 3-7 milestones, NOT one phase or one milestone), brownfield codebase mapping, phase management, session pause/resume via STATE.md, milestone audit/completion, and routing via start-next-phase. Skip for one-off features that fit a single brainstorming → writing-plans → executing-plans cycle without milestones.

Use on .NET codebases when MemoryLens MCP tools are available, in either of two paths — (1) the user directly reports a memory issue ("memory leak", "OOM", "out of memory", "high GC pressure", "process memory keeps growing", "profile memory", "investigate retained objects"), or (2) systematic-debugging is active and the symptom is memory-related. Provides snapshot/analyze/compare_snapshots over dotMemory for leak detection and before/after fix validation. Skip for non-.NET memory issues, non-memory bugs, or when MemoryLens MCP tools are unavailable.

Use on .NET / C# codebases (.cs/.csproj/.sln/.slnx) when roslyn-codelens MCP tools are available, in either of two paths — (1) the user directly asks for semantic .NET intelligence ("find callers of", "find implementations of", "who uses this", "show type hierarchy", "find unused symbols", "find circular dependencies", "what does this method do"), or (2) any superpowers skill is active (brainstorming, refactor-analysis, writing-plans, executing-plans, subagent-driven-development, systematic-debugging, test-driven-development, receiving-code-review, requesting-code-review, verification-before-completion, pre-push-review) and would benefit from grounded code-graph evidence. Applies whenever Claude would otherwise Grep/Glob for C# symbols or run dotnet build for diagnostics. Provides 32 Roslyn-powered tools (find_callers, find_implementations, get_diagnostics, get_type_hierarchy, analyze_data_flow, etc.). Skip for non-.NET projects or when roslyn-codelens MCP tools are unavailable.

Use when the user runs squad commands (/squad-init, /squad-status, /squad-sync, /squad-ask), uses `@`-mention triggers for a named specialist (`@lead`, `@backend`, `@frontend`, `@tester`, `@scribe`), asks to "activate the team" / "set up agents" / "who is on the squad", or begins brainstorming/planning on a project that already contains a `.squad/` directory. Dispatches Lead/Backend/Frontend/Tester/Scribe specialists as parallel Task subagents — each runs in an isolated context window with its own charter and tier-1 history — and accumulates per-role project knowledge across sessions in `history.md` files. Skip for one-off questions, single-file edits, or projects without an existing squad.

Use when no design system exists yet — check for docs/design/MASTER.md first, run only if missing. Triggers on "set up the design system", "no design system yet", "establish UI conventions", "define our colors and typography", or "make it look like <product>" (Linear, Stripe, Notion, Apple, Cursor, Figma, …). Also triggers implicitly when ui-workflow ui-phase needs a contract but docs/design/MASTER.md is absent. Skip if docs/design/MASTER.md already exists — update it manually instead.

Use to capture or recall cross-cutting project decisions (architecture choices, naming conventions, library selections, "going forward we'll..." commitments). Trigger on phrases like "we decided", "let's standardize on", "going forward", "this affects multiple modules", or whenever brainstorming, writing-plans, or refactor-analysis surface a decision that will outlive the current task. Also recall stored decisions at the start of brainstorming or subagent dispatch so prior choices are honored. Skip for one-off implementation choices that only apply to a single file.

Use when the user asks to "test the UI", "screenshot the page", "verify the site still works", "check the deployment", "run regression tests", or "audit the look and feel" against a web-accessible URL. Discovers and runs existing test suites (npm test, dotnet test, pytest), then performs functional checks and visual screenshots at desktop/tablet/mobile viewports via Playwright MCP, and produces a markdown report. Skip for unit/integration tests with no UI, pure backend changes, or when no URL is reachable.

Use before writing an implementation plan for a refactor that spans 3+ files or crosses module/project boundaries — codebase-wide renames, extract/move/inline operations, API surface changes, directory restructures, or interface modifications. Triggers on phrases like "refactor X across", "rename everywhere", "extract this into", "move this to", "restructure the layout", "change the public API". Maps every transitively affected file, classifies breaking vs cosmetic changes, identifies risks (dynamic references, reflection, cross-boundary impacts), and produces a safe execution order with checkpoints. Skip for single-file changes, two-file edits with no transitive impact, or pure additions.

Use for frontend phase work in two modes. (1) ui-phase before implementing — triggers on "design the UI", "spec the page", "frontend design contract", "what should this screen look like". (2) ui-review after implementing — triggers on "review the UI", "audit the implementation", "did it match the design", "compare to spec", "is the UI ready to ship". Works with any frontend tech stack (React, Vue, Blazor, Astro). Skip for pure backend or non-visual changes.

Use before pushing a branch or creating a PR. Triggers on phrases like "ready to push", "review my changes", "pre-flight check", "is this ready to merge", "about to push", "review the branch", or after the user wraps up a feature branch and signals completion. Diffs against the base branch and reviews plan adherence, code quality, commit hygiene, and regression testing, producing a PASS/FAIL verdict with a prioritized remediation plan on FAIL. Skip for in-progress work or single-commit fixes the user explicitly wants pushed without review.