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 jerseycheese/agent-skills --skill routegit clone --depth 1 https://github.com/jerseycheese/agent-skillsWrote 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/jerseycheese/agent-skills/route)<a href="https://agentmods.dev/skills/jerseycheese/agent-skills/route"><img src="https://agentmods.dev/badge/skills/jerseycheese/agent-skills/route/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/jerseycheese/agent-skills/route"><img src="https://agentmods.dev/badge/skills/jerseycheese/agent-skills/route.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 | $0.00064 | $0.01711 |
| Opus 5.5 | $0.00026 | $0.00684 |
| Sonnet 5.5 | $0.00013 | $0.00342 |
| Haiku 4.5 | $0.00006 | $0.00171 |
Grade A, and why
route 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 4d 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 — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Route
Recommend a provider and model for the task at hand. Recommend only. Never switch anything automatically, and never start the work in place of answering.
Other skills call it too: backlog-routing asks it once per batch to pick a lane.
Setup (once per machine)
The routing table and the budget tools are personal, so they live on your machine, not in this repo.
| What | Where | If it's missing |
|---|---|---|
| Routing table | $AI_ROUTING_TABLE, default ~/.agents/ai-routing.md |
Stop and say so. Start one from ai-routing.example.md next to this skill. |
| Burn reader | $AI_BUDGET_DIR/read-burn.sh --human |
Treat every seat's burn as unknown and say so. |
| Route logger | $AI_BUDGET_DIR/log-route.sh |
Skip logging and say so once. |
The two scripts are yours to write, and both are optional. The skill only needs this much from them:
read-burn.sh --humanprints one line per seat and window: provider, model or plan, window (5h or weekly), percent used, reset time, and the number's source (exact, estimate, STALE, or unknown).log-route.sh <app> <phase> <recommended> [chosen]appends one line with a timestamp to wherever you keep the log.
In a cloud session none of these exist. Ask the user for their table and a burn reading (a screenshot or pasted dashboard is fine). If they can't give one, say the call has no usage data behind it and fall back to the table's first choice for that row.
Steps
-
Read the routing table and its rules from
$AI_ROUTING_TABLE. -
Get current burn:
bash "$AI_BUDGET_DIR/read-burn.sh" --human. -
Classify the task into one phase from the table. If it spans phases, pick the phase of the immediate next step rather than the whole job.
-
Apply the ladder: name the row's first choice. Escalate only in two cases:
- the first choice's seat is at or past 90 percent used, in the window the work will actually spend
- the user says the first choice already struggled on this task
Weekly headroom on its own is never a reason to escalate.
-
Check the brief's shape. If the call lands on a seat that drops multi-issue briefs (see "Sizing a brief" below) and the work spans more than one issue, say to split it, one issue per brief.
-
Log it:
bash "$AI_BUDGET_DIR/log-route.sh" <app> <phase> <recommended> [chosen], where<app>is the app this is running in. Pass the fourth argument whenever the call departs from the row's first choice, so overridden rows surface in the log.
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.
- 4d ago Changed · +5 lines · -14 tokens per session ff12361440f3
- 23d ago First seen · 140 lines · 78 tokens per session scan A 796505c36897
route is a skill published in the GitHub repository jerseycheese/agent-skills (3 stars, last pushed yesterday), licensed MIT. It adds 64 tokens to every session and 1,711 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-09-27.
Other skills, from other repositories
local-places
Search for places (restaurants, cafes, etc.) via Google Places API proxy on localhost.
bluebubbles
Build or update the BlueBubbles external channel plugin for OpenClaw (extension package, REST send/probe, webhook inbound).
market-skill
The main marketing skill. Teaches AI how to produce high-quality marketing assets from scratch. Covers visual direction, motion design, copywriting, platform strategy, and brand identity. For Remotion videos, Nano Banana graphics, Substack headers, X posts, LinkedIn, and product launches. Stops the AI from producing…
copy-skill
Copywriting for technical founders. Covers X posts, X threads, Substack editions, LinkedIn posts, product launch captions, and README headlines. Teaches AI to write with the directness of a senior operator and the specificity of an engineer. Stops AI from producing generic marketing copy that sounds like everyone else…
nano-banana-skill
Nano Banana image generation for marketing assets. Artist-level visual direction for brand graphics, thumbnails, Substack covers, X post backgrounds, and LinkedIn headers. Covers prompt architecture, lighting direction, composition rules, style consistency, and what to never generate. Stops AI from producing generic…
notion
Notion API for creating and managing pages, databases, and blocks.