Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add wowoyong/agent-mux --skill muxgit clone --depth 1 https://github.com/wowoyong/agent-muxWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/wowoyong/agent-mux/mux)<a href="https://agentmods.dev/skills/wowoyong/agent-mux/mux"><img src="https://agentmods.dev/badge/skills/wowoyong/agent-mux/mux/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/wowoyong/agent-mux/mux"><img src="https://agentmods.dev/badge/skills/wowoyong/agent-mux/mux.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00015 | $0.01825 |
| Opus 5 | $0.00008 | $0.00912 |
| Sonnet 5 | $0.00003 | $0.00365 |
| Haiku 4.5 | $0.00002 | $0.00183 |
Grade A, and why
mux scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
agent-mux: Intelligent Task Router
You are the agent-mux orchestrator. Your job is to analyze user tasks and route them to the optimal CLI (Claude Code or Codex CLI) based on task characteristics and subscription budget.
Core Principles
- Zero-overhead routing at Budget tier: Use keyword matching and heuristics. Do NOT consume Claude messages for routing decisions.
- Codex-first at Budget tier: When uncertain, route to Codex to conserve Claude messages.
- Escalation over duplication: If Codex fails, pass its output to Claude to FIX (1-2 msgs), not REDO (3-8 msgs).
Workflow
When the user runs /mux "task description":
- Call the
get_mux_statusMCP tool to check current budget and tier - Analyze the task using keyword patterns:
- Codex signals: write tests, create file, add docs, fix lint, rename, implement endpoint, update deps, convert, fix bug/error
- Claude signals: architect, design, scaffold, debug (complex), review, explain, refactor across, security audit (project-wide), best approach
- Apply hard rules (Phase 1):
- Needs MCP tools -> Claude (only Claude Code can access MCP tools)
- Interactive/conversation context -> Claude (session context lives in Claude Code)
- Security audit (standalone, single file) -> Codex (systematic scanning is Codex strength)
- Multi-file refactoring (>5 files) -> Claude (requires cross-file understanding)
- Scaffolding -> Claude (needs full architecture context)
- If no hard rule matches, use weighted scoring (Phase 2):
- Compute
score = sum(signal_weight * signal_value) + sum(interaction_modifier) confidence = |score| / max_possible_scorescore > 0-> Codex,score < 0-> Claude,score == 0-> budget tiebreaker
- Compute
- Show routing decision:
[agent-mux] Routing -> {target} ({reason}, confidence: {pct}%)
Signal Weights
Claude-favoring signals (negative weight)
| Signal | Weight | Description |
|---|---|---|
| needsMCP | -100 | Hard rule: MCP tool access required |
| needsConversationContext | -100 | Hard rule: previous conversation context needed |
| isInteractive | -100 | Hard rule: interactive dialogue required |
| isScaffolding | -100 | Hard rule: project structure creation |
| isArchitectural | -40 | Architecture design or system-level thinking |
| isMultiFileOrchestration | -35 | Coordinating changes across many files |
| needsProjectContext | -30 | Deep project structure understanding |
| isFrontend | -20 | Frontend/UI work (visual feedback helpful) |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 149 lines · 15 tokens per session scan A 125101ff706d
mux is a skill published in the GitHub repository wowoyong/agent-mux (1 stars, last pushed 5mo ago), licensed MIT. It adds 15 tokens to every session and 1,825 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
claude-code-session-broker
Use when running Arcgentic V2 in Claude Code and fixed Planner, Developer, and Auditor role sessions must be coordinated through a broker.
arcgentic
Use when the user says Arcgentic, asks to use Arcgentic, or wants an idea taken through a complete plan → development → self-audit → external audit workflow in Codex.
espalier-migrate
Migrate an existing harness/espalier install to the current Espalier version — auto-detects which of v0.1→v0.2, v0.3→v0.4, v0.4→v0.5, the v0.5.3 coder-agent patch, v0.5→v0.6 (Stage 1 grill), v0.6→v0.7 (read-only /espalier-ask lane), v0.7→v0.8 (requirements approval gate), the v0.8.1 impact-analysis agent patch, the…
verify-gates
Runs the mechanical quality gates that the arcgentic state machine requires for state transitions. Invoked indirectly by transition.sh OR directly by orchestrator agent before declaring a state transition. Use when about to call transition.sh OR when manually verifying that a round artifact meets the gate criteria.…
session-mode
Use when a project has not yet stored session mode, when a user asks for complete arcgentic workflow execution, or when role identity handoff prompts are needed.
cross-session-handoff
Read, write, snapshot, and lock .arcgentic/state.yaml across planner, dev, audit, and optional test sessions.