createusernam
UserCross-runtime development harness for Claude Code, Codex, and OpenCode — skills, verifiable pipeline, and project templates.
Categories
Indexed Skills (32)
build-loop
Autonomous generator-evaluator cycle that builds a feature against contract.json, with the evaluator launching the live app via Playwright MCP for grading. Generator and evaluator have separate context windows. Loops until pass, restart, or max iterations. Use after /contract is written, for new features (not bugfixes — those use /diagnose + /tdd). User triggers — /build-loop, "autonomous build", "let it cycle", "generator-evaluator", "GAN loop". Requires Playwright MCP installed and contract.json present.
contract
Negotiate the "what done means" contract on disk before any autonomous code cycle starts. Produces contract.json with weighted criteria, user_flow (CJM-anchored), integrations (API contracts, data flow), restart threshold, and out-of-scope. Hard gate before /build-loop and /tdd. Use when starting a new feature, between /planning-with-files and the build cycle, or when user says "set up a contract", "what should done look like", "goal for this work", or invokes /contract. Triggers — contract, goal, "done means", acceptance criteria, rubric, user flow, integrations.
design-rubric
ONE-TIME-PER-PROJECT setup. Codifies the project's design system (Arutyunov IDS + Birman/Bureau Gorbunov) as machine-checkable criteria in design-contract.json at the project root. Future /contract invocations inherit from it; design-prototyper agent (invoked from /grill-with-docs or /build-loop) reads it before drawing. Use when starting the FIRST frontend feature for a project, when the design system changes (e.g. token rename), or when user says "design rubric", "codify the design system", "set up design contract", or invokes /design-rubric. Not for per-feature runs — re-run only when project-level design rules change.
planning-with-files
Persistent file-based planning for complex work. Creates task_plan.md, findings.md, and progress.md. Use before substantive work when the user asks to save or write a plan and execute it, says “сохрани план и выполни”, “сначала запиши план”, or “большая задача”, requests a multi-step research/build/migration/audit likely to need 5+ tool calls, or the work must survive compaction, /clear, a provider limit, or a CLI switch.
pm-review
PM gate (Phase 2-PM) — an isolated product-manager review of task_plan.md against product_brief.md before any code. Verifies journey and criteria traceability, edge-case coverage, decision rationale, risk handling, and scope. Emits pm-review.json; APPROVE required to proceed to Phase 3/4.
scaffold
Generate GRACE-marked module skeletons from the plan and contract, so the implementer gets a code few-shot instead of a spec. Writes MODULE_CONTRACT, FUNCTION_CONTRACT, typed signatures, START_BLOCK anchors, block-anchored logs and IMPL directives; leaves block bodies unimplemented. Runs with the reasoning_high binding between /to-issues and the build cycle. Use for greenfield features, not bugfixes.
grace-ontology
The GRACE annotation vocabulary — what MODULE_CONTRACT, FUNCTION_CONTRACT, BLOCK, LOG_ANCHOR and knowledge-graph tags mean, how code and architecture tiers relate, and how product intent is linked without embedding a product methodology. Read before writing or reviewing GRACE markup or extending the knowledge graph.
visualization
Render the human-track views at pipeline gates — the Mermaid plan diagram the human approves before tickets are cut (viz_before_tickets), module/dependency flowcharts, and the SUPERVISION.md index. Choose concern, then scale, then notation; never start from the tool. Separate from the agent GRACE track. Use between /judge and /to-issues or when the user asks to visualize a plan or architecture.
code-review-expert
Expert code review of current git changes with a senior engineer lens. Detects SOLID violations, security risks, and proposes actionable improvements.
design-first
Wireframe-first frontend design flow: generate wireframe → human approval HARD STOP → derive API contract → save. Invoke when user says 'design', 'wireframe', 'UI first', 'design this screen', or starts frontend work.
diagnose
Disciplined diagnosis loop for hard bugs and performance regressions. Reproduce → minimise → hypothesise → instrument → fix → regression-test. Use when user says "diagnose this" / "debug this", reports a bug, says something is broken/throwing/failing, or describes a performance regression.
grace-ask
Answer a question about a GRACE project using full project context. Use when the user has a question about the codebase, architecture, modules, or implementation — loads all GRACE artifacts, navigates the knowledge graph, and provides a grounded answer with citations.
grace-bridge
Decision guide and integration map between our Matt-Pocock pipeline and the GRACE framework. Use when starting an architecturally complex task and deciding whether to introduce GRACE XML scaffolding alongside our existing /grill-with-docs → /planning-with-files → /tdd flow. Triggers — 'apply grace', 'grace pipeline', 'should I use grace', 'grace for this task', 'hybrid pipeline', 'mix grace with our skills', 'грейс'.
grace-explainer
Complete GRACE methodology reference. Use when explaining GRACE to users, onboarding new projects, or when you need to understand the GRACE framework - its principles, semantic markup, knowledge graphs, contracts, testing, and unique tag conventions.
grace-init
Bootstrap GRACE framework structure for a new project. Use when starting a new project with GRACE methodology - creates docs/ directory, AGENTS.md, and XML templates for requirements, technology, development plan, verification plan, knowledge graph, and operational packet contracts.
grace-plan
Run the GRACE architectural planning phase. Use when you have requirements and technology decisions defined and need to design the module architecture, create contracts, map data flows, and establish verification references. Produces development-plan.xml, verification-plan.xml, and knowledge-graph.xml.
grace-refactor
Refactor GRACE-governed code safely: rename, move, split, merge, or extract modules while keeping contracts, graph, verification, and semantic markup synchronized.
grace-refresh
Synchronize GRACE shared artifacts with the actual codebase. Use targeted refresh after controlled waves, or full refresh after refactors and when you suspect wider drift between the graph, verification plan, and code.
grace-status
Show the current health status of a GRACE project. Use to get an overview of project artifacts, codebase metrics, knowledge graph health, verification coverage, and suggested next actions.
grace-verification
Design and enforce testing, traces, and log-driven verification for a GRACE project. Use when modules need stronger automated tests, execution-trace checks, or a maintained verification-plan.xml that autonomous and multi-agent workflows can trust.
grill-with-docs
Grilling session that challenges your plan against the existing domain model, sharpens terminology, and updates documentation (CONTEXT.md, ADRs) inline as decisions crystallise. Use when user wants to stress-test a plan against their project's language and documented decisions.
guide-pdf
Render an already authored Markdown research report, guide, playbook, handbook, tutorial, decision memo, or scorecard into a print-ready A4 PDF with a configurable house style. Lossless presentation only: preserves source order, headings, claims, citations, quotes, tables, and code; does not research, rewrite, restructure, add callouts, or choose narrative style. Use after `researcher` for a verification-first report or after `research-to-guide` for a reader-facing guide when users ask 'make a PDF', 'export to PDF', 'build the guide PDF', or 'print-ready version'. Detects unsupported Mermaid diagrams and stops with an explicit routing choice instead of rendering them as code.
judge
LLM-as-Judge artifact evaluation with isolated evaluator context. Supports product-brief, contract, feature, and design types. Invoke when user says "judge this", "evaluate", "validate artifact", "check quality", or after completing any phase output. Plan approval belongs exclusively to pm-review.
pipeline-status
Explain a setup-managed project's current pipeline stage, readiness, blockers, and one next action from durable on-disk state. Use when the user asks in any wording “на какой мы стадии?”, “что дальше?”, “где мы остановились?”, “продолжим проект”, “what stage are we at?”, “project status”, or returns to a project and wants orientation. Also use before resuming work when the current phase is uncertain. Do not use for the progress of one explicitly named cross-CLI task; that is workctl.
research-to-guide
Transform already validated research into a clear, evidence-faithful user guide without losing claims, limitations, sources, application, or prototypes. Owns reader contract, composition, narrative style, examples, prototype, and coverage audit. Use to rewrite, convert, expand, structure, or turn a research report, technical analysis, market study, domain review, literature review, findings file, or baseline document into a guide, playbook, handbook, tutorial, decision guide, or separate user-facing MD. Responsibility boundary: does not conduct the primary research and does not choose PDF typography/layout; use `researcher` before it and `guide-pdf` after it. Triggers: 'turn this research into a guide', 'make a user-friendly MD', 'переработай исследование в гайд', 'добавь применение и прототип', 'не потеряй важное'.
researcher
Evidence acquisition and validation for open questions: search, multi-angle or single-agent research, contradiction handling, synthesis, confidence, and judge validation. Always outputs canonical `research-state.json`; produces a verification-first `research-report.md` when the selected mode or user request requires a human-readable report. Use when users ask to research, investigate, verify, discover, analyze a market, assess technical feasibility, or fill factual gaps. Responsibility boundary: does not turn validated findings into a teaching/decision guide and does not render PDF; hand those outputs to `research-to-guide`, then `guide-pdf`.
startup
Create new project from setup template — copies templates, pre-fills config, git init, GitHub repo creation. Invoke when user says "create project", "new project", "init project", "startup", or wants to bootstrap a new workspace.
tdd
Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.
to-issues
Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices. Use when user wants to convert a plan into issues, create implementation tickets, or break down work into issues.
triage
Triage issues through a state machine driven by triage roles. Use when user wants to create an issue, triage issues, review incoming bugs or feature requests, prepare issues for an AFK agent, or manage issue workflow.
workctl
Continue one specific coding task safely across Claude Code, Codex, and OpenCode. Use when the user switches CLI, reaches a provider limit, asks for a handoff or resume prompt, returns to a repository containing multiple active tasks, or needs durable task state independent of chat history.
youtube-transcript
Extract transcripts (subtitles) from YouTube videos. Use when user needs to get text of what was said in a YouTube video, download subtitles, get captions, read video content, extract speech from video, get transcript from YouTube, or says 'transcript', 'subtitles', 'captions', 'what did they say in this video', 'extract text from video', 'субтитры', 'транскрипт', 'расшифровка видео'. Supports auto-generated and manual captions in any language.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.