add-workflow
Skill218 starsChanged 47 days agoMITView on GitHub
Write a growth.engineer workflow — workflows/<name>.md, a growth play as up to ten steps across catalog tools, with the outcome the user gets, its one motion and the inputs to ask for — and open a pull request. Use when asked to add, write, contribute or turn a growth hack, GTM play or playbook into a workflow, or to fix an existing workflow so an agent can run it.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- areal-project/AReaLtheir own version, first seen Oct 8, 2026
- inclusionAI/AReaLtheir own version, first seen Oct 8, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- inclusionAI/AReaL★ 5.9ksame name
- areal-project/AReaL★ 5.9ksame 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 repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- research-companySkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 40
Every step that calls a service names ONE published tool.
- Very specific: line 48
Read each tool's summary and notes: they say what the call returns and what it needs first (an id, a poll, a credit).
- Very specific: line 95
Check the ## Outcome against the steps: every item is something a step (or the agent's closing summary) produces.
- Distinctive: line 123
Put a key, a real email address or customer data in the file.
≈ “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
Add a workflow
A workflow is a growth play an agent can run from one file: what the user
gets, the tools in order, what to do with each, and what to ask the user for. The format is defined in workflows/README.md — read it first and copy
its template; this skill says how to write one that runs.
Input: the play, in the user's words ("when someone stars our repo, find
out who they are and reach out"), the user's GitHub login for author, and
the services they use. When they don't say which CRM, data provider or email
tool, ask; if they can't say, pick the most widely used tool the catalog has
for the job and name the choice in ## Notes.
Output: workflows/<name>.md, any missing tool files, a passing
pnpm content:check, and a pull request.
1. Name the result and the outcome
titleis the result, verb first, in 60 characters or fewer, not the method: "Email new GitHub stargazers who fit your ideal customer", not "GitHub + PDL + lemlist".summaryis one sentence of 140 characters or fewer on what it does, in order: "Finds X, does Y, and Z in<tool>." Present tense, US spelling, no words a reader can't check (meaningful, useful, thoughtful).motionis the one motion it serves, fromtags.yml.## Outcomeopens the body: one to four things the user has when the run ends (a table, drafts, records, sent emails), each one checkable and in plain words, never an input'sname. It is what people read to decide and what the agent treats as done. Never a promise of replies, meetings or revenue.- The file name is the key and the URL, forever: short, kebab-case, the play
(
github-stargazers-outbound). Checkworkflows/for one that already does this; improve it instead of adding a second.
2. Find the tools
Every step that calls a service names ONE published tool. Find them in the catalog, never from memory:
ls companies/*/tools/andgrep -l "^capability: <slug>" companies/*/tools/*.md(the capabilities are intags.yml), or over MCP at the site's/mcpendpoint:searchwith words ortags: ["capability:<slug>"], thengeta tool to read its calls andnotes. The same server'scontribute-workflowprompt hands over this guide's short form.- Read each tool's
summaryandnotes: they say what the call returns and what it needs first (an id, a poll, a credit). A step must ask only for what the call can do. - A step that relies on a filter or parameter (sign-ups since a date, deals in
a stage) names it when the tool's
notesor itsdocs:page gives it, so the running agent doesn't have to look it up. - A tool you need is missing? Add it with the
research-companyskill (from the vendor's own docs) in the same pull request. Never link a tool that does not exist, and never invent a call. - Writing a draft, picking the best match or summarizing needs no service: that
step names no tool (
3. **Write emails**. Draft …), and the agent running the file does it itself.
3. Write the steps
Up to ten, in ## Steps, each one paragraph: a bold title, with a link to
the tool's file (../companies/<handle>/tools/<name>.md, labelled with its
key, as in the template in workflows/README.md), then what to do.
- Say what to keep. End a step with "Keep …" whenever a later step needs its result — every id, email and name a later call takes. Read the steps backwards: each value a step uses must be kept by an earlier step or be an input.
- Inputs, not constants. Anything the user would change — a window, a
threshold, a channel, a campaign, a link the email points to — is an input
in
## Inputs:- `snake_case`: what it is, e.g. an example. Name every value a call needs that no step produces (a campaign's custom variable, a deal stage, an owner). - Ask in the user's words. A campaign's name beats its id when the agent can look the id up with a read-only call. When an input must be set up once first (a campaign template, a CRM property, a database), say so in its description.
- Ask before it acts. Show drafts before anything is sent. The file's Rules already say to ask before sending, spending or changing data; don't repeat them in every step.
- Triggers are runs. A workflow runs when someone runs it. For a "when X
happens" play, take a
sinceinput (the last run) so each run handles only what is new, and say in## Notesto run it on a schedule. - Cap the spend. A step that costs credits per record gets a
max_…input. - Consent. A marketing send goes only to people who opted in: filter on it, or have the step confirm it with the user.
- Say what to do with the call, not how the tool works — its setup and notes print in the file already.
4. Finish the file
- Check the
## Outcomeagainst the steps: every item is something a step (or the agent's closing summary) produces. - Header:
title,summary,author(the GitHub login),motion(onemotion:slug fromtags.yml; its header says what each means),tags(channel:entries only),addedandupdated(both today). Leavefeaturedto maintainers.
5. Check it runs
pnpm content:check # every step resolves, every tag exists, the file renders within its caps
pnpm dev # then open /workflows/<name>.md — the file an agent gets (-p <port> for another port)Then read the rendered file as the agent that will run it: list what you would ask the user for and every call you would make, step by step. Anything you would have to guess is a missing input or a missing "Keep". Fix it and check again.
6. Open the pull request
One workflow (plus any tools it needed) per pull request. Say what result it reaches and which tools it uses; the template asks how you checked it.
Never
- Link a draft or missing tool, or a call a tool's summary doesn't claim.
- Invent a company, endpoint, field or customer to make a step work.
- Put a key, a real email address or customer data in the file.
- Edit a rendered file or app code to change a workflow.
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.

