chimera-defi

Run Orbit Pilot launch workflows from Claude: CLI as source-of-truth contracts, optional browser (computer use / MCP) for manual sites, Playwright path via orbit work --playwright or orbit publish. Use when: user is launching a product, clearing the manual queue, or wants agent-driven orbit pipeline + mark-done. Packaged under apps/orbit-pilot/claude-skills/ for export with the app.

Use when a user wants to turn a rough product idea into a thoroughly planned, collaboratively authored spec bundle. This skill drives the full SpecForge workflow — Sprint Planning stages (Act 1), guided spec generation (Act 2), section-level iteration, and build-agnostic handoff. Works in CLI/TUI for solo or agent-native flows; directs to the web workspace for multiplayer collaboration.

MANUAL TRIGGER ONLY: invoke only when user types /specforge-handoff. Emits the final SpecForge handoff JSON. Packages the full export bundle (PRD, SPEC, TASKS, agent_spec.json) together with sprint planning stage provenance (completed/skipped stages and their outputs) into a single build-ready handoff.json. Build tooling is agnostic — the handoff points to no specific build tool. Use /specforge-plan for Act 1, /specforge for the full workflow.

MANUAL TRIGGER ONLY: invoke only when user types /specforge-plan. Runs the SpecForge Act 1 Sprint Planning pipeline. Walks through up to 5 optional stages (Discovery, CEO Review, Engineering Review, Design Review, Security Review), each producing a governed patch proposal against the spec document. All stages are skippable. Stage completion state is recorded in the handoff JSON for provenance. Use /specforge for the full workflow, /specforge-handoff for the final export.

MANUAL TRIGGER ONLY: invoke only when user types /specforge. Full SpecForge workflow: Act 1 (Sprint Planning) → Act 2 (Spec Generation) → Handoff. Guides you through optional planning stages (Discovery, CEO Review, Eng Review, Design Review, Security Review), then into the multiplayer spec editor with governed patch proposals, clarification queue, readiness gate, and export bundle (PRD, SPEC, TASKS, agent_spec.json, handoff.json). Also suggest adjacent skills: /specforge-plan for Act 1 only, /specforge-handoff for emitting the final handoff JSON.

Claudeception is a continuous learning system that extracts reusable knowledge from work sessions. Triggers: (1) /claudeception command to review session learnings, (2) "save this as a skill" or "extract a skill from this", (3) "what did we learn?", (4) After any task involving non-obvious debugging, workarounds, or trial-and-error discovery. Creates new Claude Code skills when valuable, reusable knowledge is identified.

Reduce total token usage across AI coding tasks by keeping discovery, reading, and follow-up context minimal. Use when file location is uncertain, the repo is large, or the user asks to explore, review, gather context, or work across multiple files. rg is the real search path; QMD is a bonus if pre-indexed. Skip for small edits with an exact file path.

Frontend development guidelines for the intelligence-exchange-cannes-web React app. Use when writing, reviewing, or modifying any file in apps/intelligence-exchange-cannes-web/src/. Encodes the design system, auth patterns, component library, and API contract for this project.

Reduce repo context cost with QMD/helper discovery, scoped rg, targeted reads, concise summaries, and AI-delegate call batching.

Use this skill whenever the task involves an Ethereum validator node — setting up a node, installing Geth/Besu/Nethermind/Reth or Prysm/Lighthouse/Teku/other clients, configuring MEV-Boost or Commit-Boost, checking node health, listing active validators, managing validator exits, BLS-to-execution withdrawal changes, or 0x02 compounding creation, operating services, cleaning node data, or updating clients. Routes all actions through `./scripts/eth2qs.sh` with strict safety guardrails. If in doubt about an Ethereum node task, trigger this skill.

Delegate bounded research, implement, debug, review, or browser tasks to Devin with workspace envelope/fallback telemetry.