README.md

---

license: other
license_name: polyform-noncommercial-1.0.0
license_link: https://polyformproject.org/licenses/noncommercial/1.0.0/
pretty_name: "token-saver - route subagents by sensitivity and capability, not habit"
language:
  - en
  - zh
tags:
  - agent-skill
  - agent-skills
  - claude-skill
  - claude-code
  - skill
  - agentskills
  - prompt-engineering
task_categories:
  - text-generation
size_categories:
  - n<1K
viewer: false
---


> **token-saver** - Route each subagent to the cheapest capable model - about 70% off execution-tier tokens in our tests.
>

> **Mirrored in three places, same version everywhere:**
> [Hugging Face](https://huggingface.co/datasets/LucioLiu/token-saver) (you are here) · [GitLab](https://gitlab.com/LucioLiu/token-saver) · [GitHub](https://github.com/LucioLiu/token-saver)
>

> **Get the files** - GitLab is the most reliable plain-git route:
>

> ```bash
> git clone https://gitlab.com/LucioLiu/token-saver.git
> 

> # or from this page
> hf download LucioLiu/token-saver --repo-type dataset --local-dir ./token-saver
> ```
>

> Then drop the folder into your agent's skills directory - `~/.claude/skills/token-saver/` for Claude Code, `~/.agents/skills/token-saver/` for Codex, or your client's equivalent.
>

> *Note: links to github.com inside the text below point to the GitHub mirror; if one does not resolve, use the GitLab mirror above - it carries the same content.*
>

> **Licence: PolyForm Noncommercial 1.0.0** - the `LICENSE` file in this repo is authoritative.

---
**English** | [简体中文](README.zh-CN.md)

# token-saver · autonomous subagent model routing 💰

> Route each subagent to the cheapest capable model that can still do the job — don't send scanning/copying work to an expensive model, and don't hand critical judgment to an underpowered one.

Which models are actually callable varies by user, account, and agent tool — and it keeps changing. token-saver doesn't maintain a vendor leaderboard and doesn't hard-code specific model names; it teaches the dispatching agent to choose based on task intensity and **what's genuinely available right now**.

> **Honest positioning**: this is a routing **discipline** (a routing guideline) — it teaches an agent how to choose and how to record it honestly, and it only takes effect if the agent follows it. It does not ship a candidate enumerator, a dispatcher, a host-log verifier, or an automatic metering tool. The ROUTE_DECISION record format and field semantics are documented in `references/route-decision.md`; the actual runtime capability must come from an adapter/tool in the current harness. Details beyond the main text live in `references/`, read on demand to control injection overhead.



## Core mechanics



- **Sensitivity-first gate**: classify the task's data as PUBLIC / INTERNAL / CONFIDENTIAL / RESTRICTED first, filter down to the models and vendors allowed to touch that level, and only then talk about capability and cost — you can't pick the cheap option first and ask afterward whether it's allowed to see the data.

- **Task-intensity intent**: light execution, routine work, deep judgment, critical review — these describe the task, not a specific model. Tiering uses a five-dimension check (uncertainty / cost of error / verifiability / context complexity / side effects), not a gut call based on the task's name.

- **Autonomous choice**: from the options the current tool genuinely allows, pick the model that's just enough to reliably get the task done.

- **Three execution modes**: `EXPLICIT` when explicit parameters are supported; `PRESET` when going through a pre-built worker; `INHERIT_ONLY` when the calling surface offers no choice.
- **Honest degradation**: inheriting the default model isn't a failure, but you may never claim it was routed by tier when it wasn't.
- **Splitting mixed tasks**: dispatch the execution segment and the judgment segment separately, to cut down on expensive-context occupancy.
- **Guarded cascade upgrade**: for tasks whose output is machine-checkable (tests / schema / compile) and whose first run can be isolated, idempotent, or fully revertible, you may run one tier down first and re-run one tier up only if verification fails; tasks with irreversible side effects (sending messages, touching a database, deploying, bulk edits/deletes) — plus critical-review-tier and non-machine-checkable tasks — are barred from cascading, to prevent "the low tier digs a hole, the high tier has to fill it."
- **Cost feedback loop**: when the host exposes usage data, feed the actual cost back as a reference for tiering choices within this session; never fabricate what isn't visible, and never let it sediment across sessions into a model ranking or price table.
- **ROUTE_DECISION dispatch record**: every routing decision logs one structured record (tier / sensitivity / mode / model chosen / cascade plan / verification status), making routing auditable and serving as an input contract for a future execution layer.

- **Quality guardrail (a three-way split of verifiers)**: deterministic verifiers (compile / test / schema) aren't a model and have no notion of "stronger or weaker"; a semantic reviewer must not be noticeably weaker than the generator; high-risk rulings must go through the main agent or a human. Never pick a weaker option for a critical result nobody will review.

- **Post-run verification**: confirm from the current platform's host logs or run metadata when needed — never trust the model's self-report, and never hard-code a log path.



## Install



**Option 1 · GitHub CLI (auto-updating)**



```

gh skill install LucioLiu/token-saver token-saver --agent claude-code --scope user

```



**Option 2 · Manual**: copy `SKILL.md` into whatever user-level skills directory your agent tool supports. Chinese-speaking users should copy [`SKILL.zh-CN.md`](SKILL.zh-CN.md) instead and rename it to `SKILL.md` locally.



## Companion blocks



- [handoff-protocol](https://github.com/LucioLiu/handoff-protocol) — the discipline for dispatching and receiving work back across multiple agents.

- [dont-reinvent](https://github.com/LucioLiu/dont-reinvent) — look for an existing skill before building one.



## Measured results (2026-07 · v1.4 evaluation)



A 30-task evaluation set (10 mechanical / 10 routine / 5 judgment / 5 mixed) × 3 dispatch strategies, run for real in a single agent-tool environment in `EXPLICIT` mode; the grader was frozen after an independent, bidirectional adversarial review (correctly failed all 30 injected wrong answers, zero false positives on 9 legitimate phrasing variants):



| Strategy | Success rate | Relative cost |

|---|---|---|

| Always inherit the main session's strong model | 96.7% | 1.00× |

| Always pin a fixed mid-tier model | 96.7% | 0.34× |

| token-saver dynamic routing (dispatch low/mid/high tier by four-level intensity) | 89.0% (mean of 3 batch runs) | 0.30× |



Honest reading:



- **Dynamic routing saves about 70% of cost relative to "always dispatch the main session's strong model directly"** — that's the real payoff under this skill's scenario assumptions.

- **On this particular task mix, "fixed mid-tier" is the best value**: the same success rate, at roughly 1/3 the cost; dynamic routing only saves an additional 12% while losing 7.7 percentage points of success. The more a task mix skews toward mechanical, bulk work, the bigger routing's payoff; the more it skews toward uniform, routine work, the more fixed-mid-tier wins.

- **Routing's losses are highly concentrated**: every one traces back to a low-tier model's off-by-one on a machine-checkable counting task — exactly the scenario "cascade upgrade" (run light first, re-run one tier up if verification fails) is meant to catch; this evaluation round did not have cascading enabled. On judgment-type tasks, routing lost zero quality (15/15 at the strong tier).



Boundaries: single platform, single round (dynamic routing n=3, everything else n=1 — variance exists); routing tiers were pre-labeled per this skill's five dimensions (the discipline's ideal execution) — how accurately an agent tiers a task on its own is a separate, untested question; `INHERIT_ONLY` / `PRESET` scenarios weren't covered; cost is a relative value against API list price. See the repo's issue tracker or the source project for how the evaluation set and grader were built.



## Version & updates



Current version **1.4.0** (see `.claude-plugin/plugin.json`).

- 1.2.0: removed vendor names, specific model names, fixed price multipliers, and a hard-coded log path — replaced with the current agent choosing autonomously based on what's actually available.
- 1.3.0: added guarded cascade upgrade and cost feedback loop, plus a near-match mapping onto existing semantic-tier aliases (fast/balanced/deep/critical, etc.).
- 1.4.0 (driven by external review): added the **sensitivity-first gate** (filter allowed models before tiering), **five-dimension tiering** (replacing gut-feel tiering based on task name), **hard preconditions for cascading** (isolated / idempotent / revertible; irreversible side effects bar cascading), the **structured ROUTE_DECISION dispatch record**, and a three-way split of verifiers (deterministic / semantic / adjudication); slimmed down the main text and moved details into `references/` for on-demand reading.



Part of the capability-block architecture behind the [Nuwa](https://github.com/LucioLiu/nuwa) digital-employee framework.



## Verification boundary



The offline contract-test runner, the component manifest and ROUTE_DECISION/policy schemas it validates, and the `SK-TS-P-001` / `SK-TS-N-001` case bundles are **not published in this repository yet** — don't expect a `tests` or `schemas` directory in what you installed. Until they ship, the field contract lives in [`references/route-decision.md`](references/route-decision.md). Nothing here has been marked PASS for semantic routing behavior: those cases stay `not-run` until an independent agent actually runs them and leaves a receipt.

## License

[PolyForm Noncommercial 1.0.0](LICENSE) — free for personal and non-commercial use.

---

> Both language editions are maintained in sync. If they ever diverge, the **Chinese edition** is authoritative — please open an [issue](https://github.com/LucioLiu/token-saver/issues) if you spot one.