ROUTING

A decision tree that chooses which coding agent should handle a request, using signals such as live incidents, protected code areas, errors, and task type.

In plain words
What is it for?
Use it to route production incidents, security or other guarded changes, stack-trace investigations, small tasks, and routine work to the appropriate agent.
Why use it?
It reduces inconsistent handoffs by applying a fixed priority order when several agents might appear suitable.

Agent

Install

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.

agentmods
npx agentmods add agents/mtvrkan/senior-dev-kit/routing
Clone the repo
git clone --depth 1 https://github.com/mtvrkan/senior-dev-kit
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,188 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.04188
Opus 5 $0.00000 $0.02094
Sonnet 5 $0.00000 $0.00838
Haiku 4.5 $0.00000 $0.00419

Measured 2d ago against content hash c0654a674870, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ROUTING 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 2d 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.

agents/ROUTING.md · 268 lines

How it starts

The opening of the file, as written. The whole thing — 268 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Routing Decision Tree

Highest-priority signal wins. Read top-to-bottom; stop at the first match.

Precedence order (memorize this line): live-incident signal (Step 0) > guard-area noun (Steps 1 and 3) > stack trace (Step 2) > too-small-to-delegate (Step 3.5) > task-type verb (Step 4). Step 0 outranks everything else because it doesn't pick an agent — it decides whether Steps 1-3 run as one coordinated, parallel dispatch instead of a single sequential match. A guard-area noun outranks every other remaining signal whenever the request changes that guarded surface — "fix CSS in the login form" is security-guard territory, not ui-fixer. A request that only references a guarded area without touching its code (writing tests against it, documenting it, researching it) routes by task type instead. Ties between two guard areas are resolved by blast radius (see "Multiple guard signals"); ties between non-guard signals by the Conflict resolution table.


Step 0 — Live incident check (before everything else)

"Prod is down" / "P1" / "outage" / "5xx spike" / "users can't log in right now" — anything reporting a live incident rather than a routine bug, even one carrying a stack trace or an obvious guard-area noun — routes to the incident-response skill first, not straight to bug-hunter or a single guard. It has no Agent tool, so it triages severity/blast-radius and produces a dispatch plan for the same Step 1/3 guards below; the calling agent/orchestrator then actually invokes them (in parallel where safe) and keeps one timeline for the postmortem. This check runs before Step 1 because it doesn't compete with the hard-stop/guard-area/stack-trace signals below — it decides whether they run as one coordinated dispatch instead of picking among them.


Step 1 — Hard stop check (always first, after Step 0)

Does the request touch any of global-CLAUDE.md's HARD STOPS nouns (auth, payment, DB schema, CI/CD, secrets, infrastructure — see that section, already loaded every session, for the exact list; not restated here so the two copies can't drift)?

Read the full file on GitHub · 268 lines

Changes

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.

  1. 2d ago First seen · 268 lines · 0 tokens per session scan A c0654a674870

Subscribe to this mod's changes

ROUTING is an agent published in the GitHub repository mtvrkan/senior-dev-kit (5 stars, last pushed 13d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 4,188 tokens. 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.