chain-launcher
Skill2 starsChanged 14 days agoMITView on GitHub
Use right after you approve a research/decision plan (a MAP) to surface the exact next command for the implement phase, making the research → approve → implement hand-off frictionless without auto-crossing the approval gate. Triggers, /chain-launcher, "what's the next goal line", "advance the chain", "fire the implement step". Never auto-fires the next phase (the gate is a deliberate human fire); does not build the prompt (that is goal-prompt).
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- apply-review-findingsSkill
- author-workflow-fanoutSkill
- delegateSkill
- design-consultationSkill
- design-fail-closed-gateSkill
- document-releaseSkill
- gistSkill
- goal-promptSkill
- grokbotSkill
- incident-as-codeSkill
- investigateSkill
- learn-captureSkill
- office-hoursSkill
- orion-deep-researchSkill
- oss-loopSkill
- pre-compactSkill
- pre-prSkill
- prove-control-bindsSkill
- prove-deploy-is-liveSkill
- readonlySkill
- reassert-guards-on-rewriteSkill
- reprobe-stale-premiseSkill
- scope-guardSkill
- shipSkill
- tdd-loopSkill
- tools-routerSkill
- triage-fanout-verdictsSkill
- xSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 53
If no implement prompt exists yet, say so and point back to goal-prompt (build it first).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Chain Launcher: surface the next command on plan approval
Overview
The compounding pattern for autonomous work is: build the prompt → run the research/decision phase → a human approves the plan → run the implement phase. The research→implement gate is a separate human fire by construction: a long autonomous run returns control only on completion, and there is no mid-run input to inject an approval into, so nothing can (or should) auto-cross the approval gate.
This skill removes the friction, not the gate: given an approved plan / prompt set, it surfaces the exact next command for the human to fire, so they don't re-derive it. Convenience over re-typing.
When to use
- Right after approving a research/decision plan (a "MAP"), to get the implement-phase command.
- Triggers:
/chain-launcher, "what's the next goal line", "advance the chain".
When NOT to use:
- To auto-fire the next phase, never; the gate is a deliberate human fire.
- To BUILD the prompt, that is
goal-prompt.
Procedure
- Identify the topic stem you used when you authored the prompt set (e.g.
timer-audit). - Locate the paired implement prompt/launcher you saved for that stem and print
its fire line verbatim: that IS the command to run. A tiny report-only helper
that globs your prompts directory for
<stem>-implement-*removes the friction; it reads and prints only, it fires nothing. - Bind to the plan you just approved, never guess on ambiguity. If more than one launcher matches the stem, do NOT print one: list the candidates and stop for the human to pick the one tied to the approved plan (matched by the plan's id / date / sha). Do not auto-select "the newest"; that is not a binding to the approval and can surface a stale or sibling launcher, exactly the failure this gate exists to prevent.
- The human reviews and fires it (editing the DO set inline first if the approval changed it).
Rails
- Never auto-fire. Print the line; the human fires it. The gate is a separate, deliberate fire.
- The helper is report-only: it reads and prints; it writes nothing and runs nothing.
- Bind to the approval. On more than one match, list the candidates and stop for human selection; never auto-pick (not even "the newest") and never fire a launcher you can't tie to the approval.
- If no implement prompt exists yet, say so and point back to
goal-prompt(build it first).
Boundary (defer to the consumer's setup)
Where you keep prompt sets, how you name the implement launcher, and which command runs an autonomous phase are your harness's conventions. This skill only insists the research→implement hand-off stays a deliberate human fire and that the next command is surfaced, not re-derived.
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.

