flame-chase
Skill3 starsChanged 18 days agoApache-2.0View on GitHub
Two chasers with different models and stances take turns on one task, each turn a fresh subagent, passing only the workspace, so every turn checks work it didn't write. Run with /baton:flame-chase "<task>".
In this repository17 other files it gives its agents.
- author-flowSkill
- cleanup-flame-chaseSkill
- cleanup-ralphSkill
- continue-loopSkill
- fixed-interruptSkill
- flowsSkill
- 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
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 41
Never add the other chaser's report, your own summary of earlier turns, check results, or hints.
≈ “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
flame-chase
Arguments: $ARGUMENTS
Load the run protocol first (Skill tool → baton:protocol; if that fails, Read ${CLAUDE_PLUGIN_ROOT}/skills/protocol/SKILL.md) and follow it. This file only adds what is specific to flame-chase.
Idea
Two agents fail in different ways. Put them on the same task, let them take turns, and give each turn nothing but the workspace. Every turn starts by checking work the other agent wrote, so one agent's blind spots don't compound over a long run. Baton makes the two chasers differ by model and by stance (see the protocol's configuration).
Roles
| Role | subagent_type | Model | Stance | Session |
| chaser-a | baton:chaser-a | chaser A model | chaser A stance | fresh every turn |
| chaser-b | baton:chaser-b | chaser B model | chaser B stance | fresh every turn |
Turn 1 is chaser-a; then they alternate: the next role is the other one than the last t line's.
Turn prompt
BATON flame-chase · turn <n>/<max> · role <chaser-a|chaser-b> · run <run-id>
TASK:
<task, verbatim>
STANCE (<stance name>):
<stance paragraph from the protocol>
You start fresh. Nothing from earlier turns reaches you except the workspace: the files and the git history of branch <branch>. Orient from them, make the most valuable progress on the TASK you can this turn, and leave the workspace so the other chaser can continue cold.Never add the other chaser's report, your own summary of earlier turns, check results, or hints. The workspace is the only handoff, and that is what makes this flow work.
Stop
- The budget is used up, or a turn fails (protocol).
- Early stop, unless
--no-early-stop: thestreak:line showsdone x2or more andempty x1or more (the last two turns, one by each chaser, both reported done, and the latest changed nothing).
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.