flows
Skill3 starsChanged 18 days agoApache-2.0View on GitHub
Lists baton's agent-loop flows and helps choose one for a task; explains the settings (models and stances), how runs use a branch, and how to resume or stop. Use when the user asks which baton flow fits a task, or what baton can do.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- erikreppel/skillstheir own version, first seen Aug 2, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- erikreppel/skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- author-flowSkill
- cleanup-flame-chaseSkill
- cleanup-ralphSkill
- continue-loopSkill
- fixed-interruptSkill
- flame-chaseSkill
- gen-ideaSkill
- gen-planSkill
- goalSkill
- parallel-chase-prSkill
- parallel-chaseSkill
- protocolSkill
- ralph-loopSkill
- recursive-lean-proverSkill
- rlarSkill
- rlcrSkill
- stateful-ralphSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Every flow runs as /baton:<flow> "<task>" [flags], works on a run branch baton/<run-id> (your branch is untouched), commits each turn, and …
- Very specific: line 41
Use a scratch repository or a clean working tree.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
baton flows
Every flow runs as /baton:<flow> "<task>" [flags], works on a run branch baton/<run-id> (your branch is untouched), commits each turn, and ends with a report on how to merge or drop the work. Shared flags: --turns N, --minutes M, --check "<cmd>", --no-early-stop, --in-place, --resume [run-id].
Pick a flow
| Flow | What it does | Use it when |
flame-chase | Two chasers with different models and stances alternate, fresh every turn; only the workspace passes between them | A long task where one agent's blind spots would compound |
cleanup-flame-chase | flame-chase plus a cleaner every few turns, NEXT.md, squashed epochs | The same, over many turns |
ralph-loop | Same task every round, fresh worker, only the repository carries over | Steady grinding on a task with a clear finish |
cleanup-ralph | ralph-loop plus the cleaner | The same, over many turns |
stateful-ralph | Same task every round, one persistent worker | Earlier reasoning matters between rounds |
continue-loop | Task once, then "continue" | The simplest long run |
goal | The task becomes the worker's own goal; it stops when it says the goal is met, with evidence | A measurable finish line, self-judged |
rlar | Persistent actor; a fresh reviewer decides "done" and writes the next prompt | "Done" shouldn't be judged by the one who did it |
gen-idea → gen-plan → rlcr | Idea draft → plan agreed by planner and fresh analyst → build with stop checks, an incremental review every round and one whole-change review by a different model | Larger features you want planned and reviewed |
fixed-interrupt | Chasers alternate every k accepted experiments scored by your evaluator; a blind review picks the winner | Optimizing a measurable score |
parallel-chase | A coordinator plans three different lanes; each lane chases in parallel; a shared leaderboard | The right approach isn't obvious |
parallel-chase-pr | Three lanes propose; merged only when a clean evaluation improves the score | Score-driven work where nothing should land unmeasured |
recursive-lean-prover | Lean 4 theorem → lemmas: a direct Lean attempt first, then prose audit and split only where needed, formalized in worktrees, wiki | Formal proofs in a Lake project |
author-flow | Writes a new flow from your description; lands only after load, fake-world and critic gates | You want a loop baton doesn't have |
Settings
Set them in /plugin → baton → configure, or in the baton rows of /config:
- chaser A: model
opus, stancebuilder(defaults) - chaser B: model
sonnet, stanceskeptic - worker model
opus, reviewer modelsonnet, default turn budget 8
Stances: builder, skeptic, simplifier, tester. Models: opus, sonnet, haiku, fable.
Before a long run
- Use a scratch repository or a clean working tree; baton refuses to start on uncommitted changes.
- Choose
acceptEditsorautopermission mode so turns don't stop on prompts. Baton never turns on bypass permissions for you. - Runs cost tokens in proportion to turns × agents, and most of a turn's cost is its agent's model. Start with a small
--turns; for cheaper runs set the chaser A and worker models tosonnetin the settings.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.