Runs a multi-model SDLC pipeline against a project brief. Routes judgment phases to Claude and mechanical phases to a cost-efficient model per policy, and reports what the run spent.
Use when a vibe-coded or hand-written codebase needs a production-readiness audit, security review, feature-completeness check, client-friendly summary, or an hours estimate. Commands: /taste (quick scan), /cook (full audit), /plate (client brief), /bill (estimate). Detects the stack (Stripe, Supabase, Convex, Firebase, Neon, MongoDB, Clerk, Expo, Capacitor, Twilio, AWS, and more) and recommends the official plugin for each.
The team in your corner — agent pipeline (spec-writer → architect → implementer → pr-reviewer → ci-fixer) plus skills, contracts, telemetry, watchdog, verified-mutations, and project scaffolding that work across iOS / Next.js / Python.
A structured feature development pipeline — brainstorm, plan, design database/API/UI, analyze edge cases, create test plans, break into tasks, and execute implementation
Reusable Claude Code skills: Sparrow static-analysis triage/autofix support, .NET diagsession/gcdump memory-leak analysis, grounded LightningChart 7.2 API assistance, frontier-model handoff prompt packaging for offline/air-gapped work, offline Windows process dump (.dmp) triage via a portable cdb/SOS pipeline, and deterministic depth-based XML-to-HTML structure reporting, a verified one-unit-at-a-time C# to modern C++17 porting loop for weak/local models, and a Claude-only plan -> develop -> QA -> wiki agent workflow with read-only reviewer subagents and build/test evidence gates, and traceable UdpSim ICD authoring with deterministic intermediate/XML validation and raw UDP payload inspection.
MCP server + skills for Agisoft Metashape Professional 2.3+ — dense reconstruction, texturing, tile export, DEM/ortho generation, photogrammetry QA, and more.
Cut token/context/cost usage in any agent or LLM workflow without degrading quality — context hygiene, output discipline, caching, batching, model routing.
Guided modernization for any legacy codebase: start with /modernize, get an assessment, an interactive map, the business rules mined from the code, and a plan you approve, then an upgrade, a rewrite or a rebuild with independent proof that the new code behaves like the old. Includes security hardening, a one-page report, specialist agents, and an optional live progress pane (early access).
Le noyau du starter kit : 6 skills génériques développeur (revue de code à divulgation progressive, stratégie de test, vérification, débogage systématique, discipline d'agent, qualité de commit) + les 5 subagents à mémoire de projet.
Lightweight PRD-first workflow with eleven user-triggered, decoupled skills: decision maps (wayfinder) for large foggy tasks, one-shot prototypes for run-to-know design questions, interview-driven PRDs, and gate-checked delivery. prd.md checkboxes track slice progress; `seed done` replays explicit checks and records done-logs; no evidence directory, task.json, or phase state machine.
Conductor skills that run the rest of the marketplace in order, with approval gates and resumable state: /feature-flow, /research-flow, /implement, /fix-flow and /upgrade build; /test-audit, /doc-audit and /harden check a branch or the whole repo; /self-review reviews your branch before a PR and /verify proves it is ready for one; /understand teaches a codebase; /learn saves what a flow taught. Also a writing skill (Williams and Bizup's Style) with a copyedit in which an Opus editor revises a document by that skill, and a code-only prose check on each Write or Edit of a Markdown file. Carries its skill packs, which only its conductors read, and installs the other plugins they use. Opt-in.