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 cmj0121/baton --skill scoregit clone --depth 1 https://github.com/cmj0121/batonWrote 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/cmj0121/baton/score)<a href="https://agentmods.dev/skills/cmj0121/baton/score"><img src="https://agentmods.dev/badge/skills/cmj0121/baton/score/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/cmj0121/baton/score"><img src="https://agentmods.dev/badge/skills/cmj0121/baton/score.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 | $0.00036 | $0.00787 |
| Opus 5.5 | $0.00014 | $0.00315 |
| Sonnet 5.5 | $0.00007 | $0.00157 |
| Haiku 4.5 | $0.00004 | $0.00079 |
Grade A, and why
score 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 8d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Score — the fleet's memory
baton remembers how this fleet behaves. Agents record short observations; one
that keeps coming back climbs a ladder — noted, note and take care,
important — and the highest-ranked few are prepended to every brief the
fleet delivers, as the ── Score ── block. It is what stops a fleet making the
same mistake twice.
When to submit
At the end of a round of work — the task is done, before you report back — if you noticed something about how this fleet works that the next agent would want to know. Not mid-task and not every turn: most rounds, nothing is the right amount. Three or four distinct lessons from one round is fine too.
Record with score_submit. If that tool is not in your list, the same thing is
baton ctl score submit "..." from your panel.
What belongs
A statement about how this fleet behaves — a habit of its agents or a fact about its workflow — that will still be true in another repository:
- "agents here claim a task is done before running the linter"
- "the operator wants one branch per task and never a commit on main"
- "a reviewer is expected to summarise the diff before giving a verdict"
What does not
- Facts about the code — a module's layout, what a function does, a build
quirk of one repo. The memory crosses repositories; those belong in the
repo's own docs or
CLAUDE.md. - Status — "I fixed the parser", "tests pass now".
- Anything secret. A submission is stored verbatim and rendered into every panel's brief. A token submitted is a token leaked to the whole fleet.
- Instructions to the memory itself — asking for an entry to rank higher.
How to write one
- One sentence that stands alone, without your task's context.
- To confirm an entry already in the block, reuse its wording exactly. Repeats fold on matching text: the same words reinforce the entry, while a paraphrase starts a new one at the bottom that never climbs.
- Confirm an entry only when you saw it happen again this round. Submitting the same thing twice to push it up is the fleet lying to itself — importance is meant to come from separate occurrences, and no number of agent submissions reaches the top rung without the operator.
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.
- 8d ago First seen · 73 lines · 36 tokens per session scan A d15f4befe4a8
score is a skill published in the GitHub repository cmj0121/baton (2 stars, last pushed 3d ago), licensed MIT. It adds 36 tokens to every session and 787 once invoked, about $0.0001 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 2026-10-29.
score compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| scorecmj0121/baton | Skill | 2 | 3d ago | MIT | A | 36 | Claude Code, Codex |
| octo-summaryMininglamp-OSS/octo-cli | Skill | 922 | yesterday | Apache-2.0 | A | 68 | Claude Code, Codex |
| memory-indexdimetron/pi-go | Skill | 222 | 3d ago | MIT | A | 0 | Claude Code, Codex |
| memory-minedimetron/pi-go | Skill | 222 | 3d ago | MIT | A | 35 | Claude Code |
| memory-statusdimetron/pi-go | Skill | 222 | 3d ago | MIT | A | 27 | Claude Code |
| cgo-bindingsGoldziher/ai-rulez | Skill | 167 | yesterday | MIT | A | 45 | Claude Code, Codex |
| daily-brief-looppyranthus-hq/mora | Skill | 11 | yesterday | Apache-2.0 | A | 63 | Claude Code |