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 agentmods add skills/everyone-needs-a-copilot/claude-copilot/protocol-routernpx skills add Everyone-Needs-A-Copilot/claude-copilot --skill protocol-routergit clone --depth 1 https://github.com/Everyone-Needs-A-Copilot/claude-copilotWrote 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/everyone-needs-a-copilot/claude-copilot/protocol-router)<a href="https://agentmods.dev/skills/everyone-needs-a-copilot/claude-copilot/protocol-router"><img src="https://agentmods.dev/badge/skills/everyone-needs-a-copilot/claude-copilot/protocol-router.svg" alt="Measured on agentmods" 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 | $0.00035 | $0.00384 |
| Opus 5 | $0.00017 | $0.00192 |
| Sonnet 5 | $0.00007 | $0.00077 |
| Haiku 4.5 | $0.00003 | $0.00038 |
Grade A, and why
protocol-router 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 yesterday.
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.
What it actually says
Protocol Router
Use this skill to classify the request and choose the right specialist flow before coding.
Classification
| Request type | Start with |
|---|---|
| bug, regression, test failure | qa |
| backend feature, refactor, architecture, performance | ta |
| user-facing feature, workflow, product experience | sd or uxd |
| physical product, connected product, hardware/software touchpoint | ind |
| visual polish, design system, component styling | uids or uid |
| auth, permissions, secrets, trust boundaries | sec in addition to primary flow |
| docs, onboarding, references | doc |
| deploy, CI, env, infrastructure | do |
Codex rule
If the user did not explicitly ask for subagents, apply the specialist lens in the main session.
If the user explicitly asked for delegation or parallel work, you may spawn subagents after deciding:
- what the immediate local step is
- which work can safely run in parallel
Standard flow
- classify the request
- identify the specialist sequence
- check whether a
tcPRD or task exists - create missing task records if the work is substantial
- execute the specialist flow
- verify with
qabefore closing implementation work
Routing patterns
Read ../protocol/references/generated-workflows.md for the catalog-derived
sequences. This router classifies requests; it does not maintain another copy
of the workflow graph.
References
Read references/routing-matrix.md when you need the full mapping.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- yesterday First seen · 50 lines · 35 tokens per session scan A e085dd327f23
protocol-router is a skill published in the GitHub repository Everyone-Needs-A-Copilot/claude-copilot (13 stars, last pushed yesterday), licensed MIT. It adds 35 tokens to every session and 384 once invoked, about $0.0002 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 2026-09-04.
Other skills, from other repositories
ship-first
Final task completion protocol: report → user-note → deploy → smoke test → close? → guide? → routing.db → propagate → STATUS → sessions. Invoked from fast-track and pipeline flows after task execution, not directly by the user. Use when: invoked after the execute step of a fast-track or pipeline task.
ui-ai-first
Final audit of a large task before closure — finds which operations are available only via code / curl / SQL and decides per each: automate with a skill or AI agent (A) or create a UI task (B). Protects against invisible usability debt. Walks through each implemented block: reads task.md + reports + guides + code →…
cadence-first
Meta-orchestrator: assigns skill-chain cadence per block (Tier 1/2/3) via Q1-Q6 rules. Reads target taskblocks + task.md, writes cadence-decisions-{R}.md artifact. Standalone (executor invokes) or Batch (generator hand-off). Use when: deciding cadence for a batch of blocks, generator hand-off from decomposition skills.
decision-first
Makes an architectural / project / scope decision using a 5-part model INSTEAD of asking the user. Structure: 🎯 Decision / Why / 🛡 Security / 📈 Scalability / Alternatives / Plain-language analogy. 1 question = 1 atomic artifact. Use when: the agent is about to ask an architectural / scope question…
library-first
Mandatory protocol before executing any fast-track task. Analyzes the task, builds a table: what we do / where it comes from / how many lines of code. Principle: maximum reuse of existing libraries and components, minimum new code. Waits for explicit user approval — does nothing until confirmed. Use when: fast-track…
fixture-new
Creates a parity fixture — the frozen scenario plus the contract its output must satisfy. Asks which skill and case, what shape the run must produce, and writes input.md and expect.yml. Ends by proving the new fixture actually fails on an empty directory. A fixture that passes when nothing ran is worse than no…