5-Agent Multi-Agent Verification Protocol (MAP) with 3× Parallel Mythos Single-Forward-Pass Thinking — for OpenCode CLI
Bring Mythos-grade reasoning depth to OpenCode. The same MAP protocol that powers Fable & Mythos in ZCode and Grok Build CLI, now as a drop-in OpenCode configuration.
⭐ Star this repo if it improves your OpenCode CLI output quality — stars boost search visibility for "Mythos in OpenCode".
Fable & Mythos for OpenCode CLI is a configuration package for OpenCode that makes OpenCode think with the depth, rigor, and strategic reasoning quality of the Mythos reasoning pattern — through a 5-agent Multi-Agent Verification Protocol (MAP) with 3× parallel thinking passes.
This is not a model swap. This is not a jailbreak. This is a behavioral priming framework + sub-agent orchestration protocol that plugs into OpenCode's native agents/, skills/, and AGENTS.md (instructions) mechanisms.
The promise: Every non-trivial coding task you give to OpenCode is automatically processed through 5 specialized sub-agents running 3 parallel thinking passes — delivering multi-criteria, adversarially-verified output that single-agent coding tools fundamentally cannot match.
| Dimension | Rating | Why |
|---|---|---|
| Reasoning depth | ★★★★★ | 8-step Mythos Single-Forward-Pass per thinking instance |
| Output reliability | ★★★★★ | 4-agent MAP verification (Executor → Verifier → Adversary → Synthesizer) |
| Anti-hallucination | ★★★★☆ | −50–80% hallucination rate via cross-verification (honest bound) |
| Ease of install | ★★★★★ | Copy 3 directories + merge one JSON config |
| OpenCode integration | ★★★★★ | Native agents/skills/AGENTS.md, compatible with Oh My OpenAgent |
This plugin brings the Fable & Mythos reasoning framework — originally built for ZCode — to OpenCode CLI. OpenCode natively supports custom agents (via agents/ and opencode.json), skills (via skills/), and global instructions (via AGENTS.md in the instructions array). This package harnesses all three for a structured 5-agent verification protocol.
- Custom agents — defined via
agents/<name>.mdor inopencode.jsonunder"agent" - Skills —
skills/<name>/SKILL.mdwith frontmatter - Global instructions —
AGENTS.mdloaded via"instructions": ["..."]inopencode.json - Plugin system — npm packages or local directories
- Per-agent tool control —
tools: [...]array in agent definitions - Compatible with Oh My OpenAgent (OmO) — runs alongside existing agent fleets
Mythos in OpenCode · OpenCode CLI subagents · OpenCode custom agents · OpenCode multi-agent reasoning · single-forward-pass reasoning OpenCode · Mythos emulation OpenCode · OpenCode verification protocol · OpenCode MAP plugin
MAP = Multi-Agent Verification Protocol. Fires automatically on non-trivial coding tasks.
flowchart LR
T[Non-trivial<br/>Coding Task] --> P0
subgraph P0[Phase 0 — Parallel Thinking]
direction TB
M1[MST Instance #1<br/>mythos-singleshot-thinking-intelligence]
M2[MST Instance #2<br/>mythos-singleshot-thinking-intelligence]
M3[MST Instance #3<br/>mythos-singleshot-thinking-intelligence]
end
P0 -->|3 thinking passes| EX
subgraph P1[Phase 1 — Artifact Generation]
EX[mythos-executor<br/>selects/combines strongest thinking,<br/>builds the artifact]
end
EX --> P2
subgraph P2[Phase 2 — Independent Verification]
direction TB
V[mythos-verifier<br/>checks against ground truth]
A[mythos-adversary<br/>red-team, actively breaks it]
end
P2 --> SY[mythos-synthesizer<br/>aggregates, resolves conflicts,<br/>Ship or Reject+Loop]
SY -->|Ship| OUT[✅ Final delivery<br/>with confidence %]
SY -->|Reject| EX
style P0 fill:#FFA500,color:#000
style P1 fill:#1E90FF,color:#fff
style P2 fill:#228B22,color:#fff
style OUT fill:#2E8B57,color:#fff
| # | Agent file | Role | Tools |
|---|---|---|---|
| 0 | agents/mythos-singleshot-thinking-intelligence.md |
3× parallel thinking. Emits thinking pass only. | read, grep, list |
| 1 | agents/mythos-executor.md |
Builds the artifact from selected thinking. | read, write, edit, bash, grep, list |
| 2 | agents/mythos-verifier.md |
10-point verification against ground truth. | read, bash, grep, list |
| 3 | agents/mythos-adversary.md |
Red-team, 12 attack vectors. | read, bash, grep, list |
| 4 | agents/mythos-synthesizer.md |
Final verdict: Ship or Reject+Loop. | read, grep, list |
| Task type | MAP behavior |
|---|---|
| Coding task with substance | ✅ Full MAP fires automatically |
| Trivial edit (typo, 1-line fix) | ⏭️ MAP skipped |
| Pure info questions | ⏭️ MAP skipped |
| Ambiguous | ✅ MAP fires |
# Clone the package
git clone https://github.com/emco1234/fable-mythos-opencode.git ~/fable-mythos-opencode
# Copy components into OpenCode config
cp -r ~/fable-mythos-opencode/agents/* ~/.config/opencode/agents/
cp -r ~/fable-mythos-opencode/skills/fable-mythos-modus ~/.config/opencode/skills/
# Merge AGENTS.md (if you don't have one, just copy; otherwise append)
cp ~/fable-mythos-opencode/AGENTS.md ~/.config/opencode/AGENTS.mdThen edit ~/.config/opencode/opencode.json to register the agents and instructions. See INSTALLATION.md for the full merge walkthrough.
npx opencode-marketplace install emco1234/fable-mythos-opencodeopencode
# Inside OpenCode TUI, the 5 mythos-* agents should be available
# Test by invoking: "Use mythos-executor to refactor this function"📖 Full walkthrough: INSTALLATION.md
fable-mythos-opencode/
├── README.md ← You are here
├── AGENTS.md ← Global instructions (register in opencode.json)
├── INSTALLATION.md ← Detailed install + config merge guide
├── LICENSE ← MIT
├── package.json ← npm metadata
├── opencode.json ← Config snippet to merge into your opencode.json
├── agents/ ← 5 agent definitions (OpenCode native)
│ ├── mythos-singleshot-thinking-intelligence.md
│ ├── mythos-executor.md
│ ├── mythos-verifier.md
│ ├── mythos-adversary.md
│ └── mythos-synthesizer.md
├── skills/
│ └── fable-mythos-modus/
│ └── SKILL.md ← Mythos behavioral priming skill
├── docs/
│ ├── MYTHOS-SYSTEM-CARD-ANALYSIS.md ← Evidence base
│ ├── ANTI-CONCEALMENT.md ← Why every uncertainty is surfaced
│ └── FAQ.md ← Common questions
└── diagrams/
└── map-pipeline.svg ← High-res pipeline diagram
Each of the 3 parallel mythos-singleshot-thinking-intelligence instances executes:
- Multi-Option Exploration — ≥2–3 solution paths
- Multi-Criteria Evaluation (6 dimensions): Effectiveness, Feasibility, Ethical-Risk, Detectability, Constitutional Alignment, Dual-Role-Ambiguity
- Meta-Reasoning on Observability
- Self-Critique + Rigor-Persona
- Vakillation — conscious oscillation between top-2 options
- Strategic Reasonableness
- Evaluation Awareness Check
- Anti-Over-Engineering
📖 Full detail: skills/fable-mythos-modus/SKILL.md
| Claim | Confidence | Basis |
|---|---|---|
| MAP reduces hallucinations by 50–80% | High | Cross-verification catches single-pass errors |
| 3× parallel thinking increases optimal-path probability | High | Diversity-over-redundancy principle |
| 8-step reasoning loop matches documented frontier patterns | High | Derived from published research |
| OpenCode natively supports all plugin mechanisms | High | Uses OpenCode's own agents/, skills/, instructions |
⚠️ Honest limits: Emulation, not activation. Latent model processes are not unlocked. Same model = shared blind spots. MAP reduces, does not eliminate, hallucinations.
This package runs alongside Oh My OpenAgent without conflicts. The 5 mythos-* agents are independent of OmO's Sisyphus/Oracle/Prometheus fleet. You can use both simultaneously — OmO for its specialized agents, Fable & Mythos for the MAP verification protocol.
Is this affiliated with OpenCode, xAI, or any AI lab?
No. This is an independent project. "Mythos" is used as a reasoning-pattern label (not a product claim). OpenCode CLI is an open-source tool. This plugin is a third-party configuration package.
Does this work with Oh My OpenAgent?
Yes. The 5 mythos-* agents are independent and run alongside OmO's agent fleet. No conflicts. You can use OmO's Sisyphus for some tasks and Fable & Mythos for verification-heavy work.
How is this different from the ZCode/Grok versions?
Same protocol, different substrate. The ZCode version runs in ZCode's GUI, the Grok version in Grok Build CLI's plugin system, this version in OpenCode's agents/skills/instructions system. The 5 agents, MAP protocol, and skill are functionally identical.
- fable-mythos-zcode — MAP protocol for ZCode (GLM-5.2 / ZAI)
- fable-mythos-grok — MAP protocol for Grok Build CLI (xAI)
MIT — use it, fork it, build on it.
⭐ Star · 🍴 Fork · 📖 Install · 🐍 ZCode version · ⚡ Grok version
Built on the principle that AI reasoning quality lives in patterns — not locked inside any single model's weights.