forge-idea-framing
Skill0 starsChanged 3 months agoMITView on GitHub
Turns a vague idea into a framed, testable and deliberately small problem for a solo developer. Produces docs/00-framing.md covering the problem, the user, the current alternative, the v1 scope, what is explicitly out of scope, and the kill criteria. Use this skill whenever someone shows up with "I have an idea for a project", "I want to build an app that…", "is it worth building…", or asks for an MVP, requirements or a starting point — including when the idea already sounds clear, because that is exactly where scope goes wrong.
In this repository6 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Not a feature list. Not a framework choice.
- Very specific: line 58
Write docs/00-framing.md, then stop and ask for review.
- Distinctive: line 111
If framing reveals that the problem doesn't exist, that an existing tool already solves it, or that the time budget is an order of …
≈ “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
Idea framing
Goal: leave this step with a problem, a user, and a boundary. Not a feature list. Not a framework choice.
An unframed idea has one telltale symptom: you can add anything to it without it looking less coherent. Framing is what makes additions expensive.
Order of questions
One question per message, multiple choice where possible. Don't move on until the previous answer is sharp. The order is not arbitrary — each answer narrows the space of the next.
- Who has the problem? A concrete person, not a segment. "Me, when I prepare audits" beats "security officers at small companies."
- What does that person do today instead? There is always a current solution — a spreadsheet, a notebook, a script, or giving up. If you can't find it, the problem may not be felt.
- What's wrong with that current solution? This is where the value lives. If the answer is "nothing serious", say so plainly.
- What does success look like, measured? An observable number, not "it works well."
- Which constraints are non-negotiable? Budget, hosting, data sovereignty, offline operation, language, regulation, target devices.
- How many hours per week, and until when? This answer determines v1 scope more than any other.
- What already exists? Repository, data, accounts, available infrastructure.
The ending question, asked at the beginning
"By what date, and on what signal, do you decide to stop if this doesn't take off?"
Most people have not thought about it. It is the most useful question in the whole framing. A solo project without kill criteria doesn't stop — it spreads and occupies indefinitely.
Help phrase something observable and dated: "if in eight weeks I'm not using it myself three times a week, I stop."
Cutting scope
Once the problem is understood, propose a deliberately frustrating v1: the smallest thing that solves the problem for the person identified in question 1.
Method: list everything the human mentioned, then for each item ask "if I remove this, is the main problem still solved?" If yes, it moves to Out of scope — not to the bin. Writing down what you are not doing is what prevents drift; an unlisted idea comes back through the window three weeks later.
Expect resistance. Argue on cost, not on principle: "that adds a table, a screen and an error case — two weekends. Is the problem not solved without it?"
Artifact
Write docs/00-framing.md, then stop and ask for review. Full template in
references/templates.md; expected structure:
# Framing — <project>
Date: YYYY-MM-DD · Status: proposed
## Problem
<2-4 sentences. The problem, not the solution.>
## Person affected
<Who, in what situation, how often.>
## Current solution and its limits
<What is done today and why it falls short.>
## What changes if this works
<The benefit, from the person's point of view.>
## Measurable success
<Observable indicator + threshold + date of observation.>
## Non-negotiable constraints
- <budget, hosting, data, language, offline, regulatory…>
## Real budget
Time available: <h/week> · Horizon: <date> · Money: <amount/month>
## v1 scope
- <3 to 7 capabilities, in user language>
## Out of scope (explicitly)
- <what you are not doing, and why>
## Kill criteria
<Observable signal + decision date.>
## Open questions
- [NEEDS CLARIFICATION: …]Self-review before presenting
Reread with fresh eyes and fix in place:
- Any
TBD, "to be defined", or vague requirement left? - Does the v1 scope fit the declared time budget? Be honest about the order of magnitude.
- Is success measurable, or is it an intention disguised as a metric?
- Does any section contradict another?
- Does the document describe a problem, or has it drifted into a solution?
When to say no
If framing reveals that the problem doesn't exist, that an existing tool already solves it, or that the time budget is an order of magnitude below the scope, say so. This is the only moment in the route where it costs nothing. Then offer: cut, buy instead of build, or abandon. The human decides; your job is to make the trade-off visible.
Exit
Once approved, the route continues to the constitution skill (if the project is new) then to specification. Invoke no other skill.
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.

