new-motion
Skill12 starsChanged 18 days agoMITView on GitHub
Scaffold a new GTM motion end to end across all three layers. Use whenever creating a new motion, campaign, play, or automation in this repo, so the narrative (context), the implementation (infra), and the regression suite (evals) ship together and never drift apart. Enforces the outreach hypothesis rule.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .cursor/rules/manifest-guardrails.mdcCursor rule
- llms.txtllms.txt
- cadenceSkill
- capture-feedbackSkill
- context-hygieneSkill
- new-executionSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 29
Only outreach motions are campaigns.
- Very specific: line 65
Eval third: If the motion calls an agent or an ai() step whose prompt you wrote, add evals/<name>/ with representative cases and wire it …
- Very specific: line 69
Verify: Run npm run lint, then open the PR.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
New motion, three layers at once
A motion is not just the automation. It is the narrative, the code, and the guardrail. Create all three in one PR.
On vocabulary: a motion is the go-to-market move, and its narrative
lives in context/motion/. A play is a specific Cargo resource
(definePlay: an automation that watches a data model and runs per row). A
motion is implemented as a Cargo play, a Cargo tool, or an agent, depending on
how it is triggered. Do not confuse the two.
Step 0: is this a campaign?
Set kind: in the narrative's frontmatter. It changes what you must declare.
| kind | What it is | Must declare a hypothesis? |
outreach | A campaign: it sends messages to humans | Yes |
enrichment | Fills in data | No |
scoring | Ranks records | No |
routing | Assigns records to people | No |
signal | Detects a trigger | No |
Only outreach motions are campaigns. The others declare their intent through the
outcome they serve in plan/outcomes.md.
Steps
Context first. Copy
context/motion/_template.mdtocontext/motion/<name>.md. Fill in the trigger, the sequence, the owner, the success metric, andserves:(the outcome inplan/outcomes.mdthis makes real). Reference the ICP and personas with[[icp/<slug>]]and[[persona/<slug>]]wikilinks.If
kind: outreach, write the hypothesis before a single line of copy exists: segment, angle, trigger, expected_reply_rate. This is what lets the result be judged against a bar set in advance rather than one invented afterwards. The linter fails the PR if it is missing.Outreach motions may cite only
validatedorprovenknowledge. If the angle leans on a claim that is stillconfidence: hypothesis, there are two honest options: validate the claim first (see thecapture-feedbackskill), or change the angle. Never upgrade a claim just to unblock the copy.Infra second. Pick the implementation by how it is triggered:
- Watches a data model, runs per row:
definePlay+defineWorkflow. - Runs on demand, or called by an agent:
defineTool+defineWorkflow. - Reasons, decides, and calls several tools:
defineAgent.
The file goes in
infra/as<name>.ts.infra/is flat by default: only nest (infra/plays/,infra/tools/,infra/agents/) once a use case has enough files that finding one is work, or because a cookbook you installed already nests. Checkinfra/README.mdfor what the directory looks like now before you decide.Workflow bodies are parsed, not executed: no await, no try/catch, no closures, no destructuring inside the body. Reference existing resources by imported handle, never by pasted UUID. Run
npm run typecheck.- Watches a data model, runs per row:
- Eval third. If the motion calls an agent or an
ai()step whose prompt you wrote, addevals/<name>/with representative cases and wire it intoevals/promptfooconfig.yaml. - Verify. Run
npm run lint, then open the PR. Confirm the cargo plan comment shows only the resources you intended to add. The narrative and the plan diff should tell the same story.
Naming
Kebab-case, verb-first where possible: inbound-router, churn-winback,
signal-to-sequence. Same slug across all three layers.
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.

