define
/define
Write or update intent.md for a Manykind workflow or set of workflows, the problem-first "why" file with Goal, Outcome, Trigger, Done when and Examples, before design.md. It works from an existing workflow's code (manykind.config.ts) or a console (ROC) export, free text ("I want a workflow that…"), a Slack thread or a meeting transcript. Use when someone wants a new workflow or set, changes what a workflow is for, its outcome or its trigger, or asks for an existing workflow to be documented. Not for reviewing, hardening, bug-hunting or small fixes ("is it production-ready?", "find the bugs"):
- Author
- manykind-ai ↗
- Repository
- plugin ↗
- Licence
- MIT
- Last repository change
- 10.11.2026
Assessment and security
- Files in this package
- 3
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 10.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Write or update intent.md for a Manykind workflow or set of workflows, the problem-first "why" file with Goal, Outcome, Trigger, Done when and Examples, before design.md. It works from an existing workflow's code (manykind.config.ts) or a console (ROC) export, free text ("I want a workflow that…"), a Slack thread or a meeting transcript. Use when someone wants a new workflow or set, changes what a workflow is for, its outcome or its trigger, or asks for an existing workflow to be documented. Not for reviewing, hardening, bug-hunting or small fixes ("is it production-ready?", "find the bugs"):
Triggers from the description
Original author description
Write or update intent.md for a Manykind workflow or set of workflows, the problem-first "why" file with Goal, Outcome, Trigger, Done when and Examples, before design.md. It works from an existing workflow's code (manykind.config.ts) or a console (ROC) export, free text ("I want a workflow that…"), a Slack thread or a meeting transcript. Use when someone wants a new workflow or set, changes what a workflow is for, its outcome or its trigger, or asks for an existing workflow to be documented. Not for reviewing, hardening, bug-hunting or small fixes ("is it production-ready?", "find the bugs"):
How it works
- Where the file goes
- Sources, most useful first
- Process
- Writing rules
- Sets of workflows
- When intent.md changes
SKILL.md preview
---
name: define
description: Write or update intent.md for a Manykind workflow or set of workflows, the problem-first "why" file with Goal, Outcome, Trigger, Done when and Examples, before design.md. It works from an existing workflow's code (manykind.config.ts) or a console (ROC) export, free text ("I want a workflow that…"), a Slack thread or a meeting transcript. Use when someone wants a new workflow or set, changes what a workflow is for, its outcome or its trigger, or asks for an existing workflow to be documented. Not for reviewing, hardening, bug-hunting or small fixes ("is it production-ready?", "find the bugs"): those get a findings list and no intent.md.
---
# Define a workflow: intent.md
This skill writes the **why** of a workflow into `intent.md`: the problem
first, before anything about steps or storage. The **how** goes in
`design.md` (skill `design`), as the next step. The concepts and the two-file
rule are in the `workflows` skill.
**Only when it's wanted.** Run this for a new workflow or set, a change to
the goal, the outcome or the trigger, or a request to document a workflow. A
review, a production-readiness check, a bug hunt or a small fix doesn't
create `intent.md`; it gets a ranked findings list (the `workflows` skill,
"Shape the output by the request").
Your knowledge of Manykind may be out of date, so **prefer retrieval**. To read
an existing workflow, use its files and `mk schema workflow`, or
`node_modules/@manykind-ai/agent-sdk` types. For concepts, fetch one docs page as
the `workflows` skill's `references/docs.md` describes. Don't rely on memory.
## Where the file goes
- **A workflow in code:** next to its `manykind.config.ts`.
- **A console-only workflow:** in a local folder named after the workflow.
Ask once where. If the user has no preference, use `./workflows/<id>/`.
- **A workflow in a set** (several workflows that share data and one
outcome): one `intent.md` for the set, in the parent folder. The workflows
in it get no `intent.md` of their own. See
[Sets of workflows](#sets-of-workflows).
- **The file already exists:** update it in place. Change only the sections the
new information affects. Keep every other section, comment and extra
heading.
Template: [templates/intent.md](./templates/intent.md). Use its `##` headings
exactly. Users may add more sections, and you never remove them.
## Sources, most useful first
1. **An existing workflow** (the main path):
- In code: `manykind.config.ts` holds `goal`, `card`, the trigger nodes, the
`output` Result and the edges to it. Also read the README and the
comments.
- From the console: the exported graph JSON holds the same fields.
- If the CLI and the access are available, these show what it
really does:
- `mk vault inspect --vault <id> --agent <workflowId>` for its triggers,
schedules, connections and versions;
- `mk runs list --workflow <id>` and `mk runs get <runId>` for its
recent runs.
Describe what the workflow **does today**. Where the code and the user's
description differ, write down both and ask which one is the intent.
2. **Conversation material.** From a Slack thread, a transcript or notes, pull
the goal, who gets the outcome, what starts it, and any examples people
gave.
3. **Free text from the user.**
## Process
0. **New workflow or a change to this one?** If the request adds a different
trigger or cadence (for example a daily summary on a workflow that runs per
batch), it is usually a new consumer workflow that reads this one's data.
Say so, and write a new `intent.md` for it. The existing workflow's
`design.md` then only gains a "Read by" line, plus any data the consumer
needs that isn't stored yet. If the request describes several workflows
that only make sense together, write a set-level file first (below).
1. Read every source you were given. Draft **every section you can** from them.
2. List the sections that are still empty or uncertain. Ask about them **one
question at a time**, in plain words. Ask only about Goal, Outcome, Trigger
and Done when. Ask three questions at most. If something is still unknown
after that, leave the section out or put the question under one
`## Open questions` heading, and move on. Never write `TBD` stubs.
3. **Examples** are optional at this stage. Fill them from the sources:
- an existing workflow's recent runs (`mk runs get`;
`mk eval capture <dataset> <context…>` turns real runs into cases), or
its fixtures and tests;
- examples people gave in the thread or the transcript.
If there are none, leave the table empty. Tell the user that release needs
them, and that the `evaluate` skill turns them into a dataset.
4. Write the file and summarise in two or three lines what you filled in, what
you inferred, and what is still open.
5. Go on to `design` (how it works on the platform and stores data), or
suggest it if the user wants to stop here. For a workflow that already
works, its design can be reverse-engineered the same way.
## Writing rules
- Use business words. Name the outcome and who receives it. Leave out steps,
node kinds, SQL and model names: they belong in `design.md`.
- **Done when** holds checks someone can verify, from a run's Result, its data
or what a person sees. "Every alert gets one reply in its thread within 10
minutes" is a check. "Works well" is not.
- **Trigger** also says what is ignored: other sources, duplicates, the
workflow's own output.
- **Examples** pair a realistic input with its expected outcome. Use real
shapes, with sensitive values replaced.
- Don't make up facts to fill a section. Leave it out, or list the question
under `## Open questions`.
## Sets of workflows
Use a set-level `intent.md` when two or more workflows serve one outcome and
share data: a funnel, or a factory such as research → content → distribution
→ analytics. Template: [templates/set-intent.md](./templates/set-intent.md).
- The set's file holds the Goal, the Outcome, the workflows (one row each
with its trigger and what it contributes), Done when for the whole set, and
Examples.
- Per-workflow Done when items and Examples go in the same file, labelled
with the workflow's name (the examples become that workflow's dataset).
- Don't write an `intent.md` inside each workflow's folder. A workflow gets
its own only when it has its own goal that differs from the set's.
- Layout and the data checks across the set: the `workflows` skill's
`references/sets.md`.
## When intent.md changes
Update it when the goal, the outcome, the trigger or the done condition
changes. A prompt tweak, a new parameter value or a bug fix doesn't change
it. If the change also alters steps or data, `design.md` changes as well.
Files in this package (3)
- SKILL.md 7 KB
- templates/intent.md 1 KB
- templates/set-intent.md 1 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · manykind-ai/plugin | 100 | MIT |
| doodledood/manifest-dev | 100 | MIT |
| SensorsIot/Universal-ESP32-Workbench | 100 | MIT |
| Trevoke/org-gtd.el | 100 | GPL-3.0 |
Database snapshot: 13.11.2026 · package version dfbe83d6bc3d (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100workflows
Background knowledge for building on Manykind. Covers what a workflow is (triggers, steps, gates, runs, Result), where its data lives (cubbies, migrations, shared workspaces, the memory bank), how workflows build on each other's data through shared keys and events, the two files… - 100develop
Write, extend and debug Manykind workflow and agent code. Covers defineWorkflow graphs (steps, edges, gates, Result), agent engagements and event handlers, cubbies and SQL migrations, the memory bank, widgets and tests, then validating, building and running a workflow once… - 100write-intent
Use when asked to write, draft or start an intent document for a change, before any design or code exists. - 100naming-workflows
This skill should be used when the user wants to name a workflow, write workflow descriptions, standardize workflow documentation, add a workflow to the AI Registry, or structure workflow entries. Generates consistent, outcome-focused names and descriptions for business… - 100goal-draft
Turn a fuzzy intention into a /goal objective the Goal verifier can actually judge - one outcome, numbered binary "Done when" checks that leave evidence in the transcript, guardrails, a budget, and a block protocol. Use when the user wants to set or define a goal, asks whether a… - 95define-goal
Help the user define a concrete, measurable goal before starting work, especially when they ask to use the goal tool, create a goal, set an objective, clarify success criteria, or turn a fuzzy intention into a quantitative outcome. Use this skill for goal creation and goal…