manusco/resonance

> Operator-grade AI agent skills for builders. A cross-tool skill library and slash-command system for Claude Code, Cursor, Codex, and opencode, covering strategy, design, engineering, marketing, sales, and ops.
Claude Code  ·  Cursor  ·  Codex  ·  opencode  ·  Antigravity
---
What it is
Resonance is an AI agent skill library you drop into any project. It turns a general coding agent into a roster of specialists that follow the same expert protocol every time, on whatever tool you use.
- **73 domain-tested skills** across design, engineering, finance, leadership, marketing, ops, people, research, sales, software, strategy, and success. Each skill is a structured procedure with prerequisites, a step-by-step algorithm, a Recovery path, and a Definition of Done, backed by a deep reference library. Not a prompt. A protocol.
- **38 slash commands** like /brief, /plan, /grill, /council, /build, /debug, /design, /test, /improve, and /ship. Type the command, or describe the job and let the specialist auto-fire.
- **Cross-tool by design.** One source compiles to the native format of every major agent tool. The SKILL.md / AGENTS.md open standard is the shared content; the Forge emits the per-tool command shims and the per-tool context bridge, so the operating standard, the commands, and the project memory all load after a clone in Claude Code, Cursor, Codex, opencode, and Antigravity.
- **A project memory** (.resonance/) that loads at the start of every session and the agent writes back to. It does not forget your architecture, your decisions, or your voice.
- **Token-efficient.** The shared operating standard is stated once in AGENTS.md, not repeated in every skill. Compiled skills are lean, so per-session context stays cheap.
You get consistent, high-quality output because the agent runs the same protocol every time, not because you remembered to ask nicely.
---
Works in your tool
| Tool | How commands are delivered | Status |
| :--- | :--- | :--- |
| **Claude Code** | Native skills in .claude/skills/ | /plan, /ship, ... |
| **Cursor** | Skills in .cursor/skills/ | /plan, /ship, ... |
| **Codex** | Agent Skills discovery plus AGENTS.md routing | describe the job or select a skill |
| **opencode** | Commands in .opencode/commands/ plus AGENTS.md routing | /plan, /ship, ... |
| **Antigravity** and other AGENTS.md tools | AGENTS.md command map | describe the job |
The command shims are generated from one source (.forge/commands.json) by the Forge, which also emits the per-tool **context bridge**: the file each tool loads at session start, pointing it at AGENTS.md and the .resonance/ memory. Claude Code loads CLAUDE.md, not AGENTS.md, so the Forge writes a root CLAUDE.md that imports both; Cursor gets an always-applied .cursor/rules/resonance.mdc; Codex, opencode, and Antigravity read AGENTS.md natively. Without the bridge the operating standard and memory never reach the model. Adding a new tool is one host-config line.
---
Quickstart (60 seconds)
Open the folder in Claude Code, Cursor, Codex, or opencode. The slash commands are already committed, so they work immediately. In your AI chat:
That is the whole setup. No install step, no plugin required.
---
The command catalog
Every command is a structured procedure with a Definition of Done, not a loose prompt. Full map in AGENTS.md.
The registry contains **38 commands**.
**Autonomous loop**
/goal: The autonomous goal loop: frame, decompose, then build and verify each slice against real checks, bounded, never auto-ship.
**Inception**
/init: Bootstrap the .resonance/ project memory (soul, state, docs scaffold). Run once per new project. · /venture-model: Model the business, offer stack, and revenue math before planning. · /brief: Turn a rough request into an intent-faithful execution brief, then run or route it within the user's authority. · /blueprint: Create or revise a durable architecture constitution, or check a plan, change, PR, or release for architectural drift. · /plan: Turn a feature or idea into an approved implementation plan sized to the decision, with executable steps and verification. · /grill: Stress-test a plan or design before any code: relentless one-question-at-a-time interrogation to shared understanding. · /council: Challenge an analysis or high-risk decision through relevant specialist reviews, debate, scenarios, and reconciliation. · /gtm-thinker: Stress-test and expand a go-to-market campaign concept into a strategic blueprint with kill criteria. · /market-research: Discover Existential Data Points in a B2B SaaS vertical. Positioning from nice-to-have to must-have. · /update-roadmap: Sync .resonance/01_state.md with the git log so the map matches the territory.