agentleFS
gdhub-iogdhub-io/character-forge-mcp/.claude/skills/intent/SKILL.md

intent

Skill1 starsChanged 4 days agoMITView on GitHub

Start a piece of work by writing its intent — what is wanted, why, what must not move, and when it is done — into a GitHub issue, then open the branch linked to it. Use when picking up an issue, when asked to plan a change before building it, or when an issue's intent needs checking against what the work has become.

16 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use hud, ralplan and 6 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • ibahgatSkill

    hud

    Control omi visual HUD profile and keep status output readable in normal, compact, or hidden mode.

    Also in 3 projects that use this

  • ibahgatSkill

    ralplan

    Create a strict, quality-first execution plan with explicit gates and rollback-ready sequencing.

    Also in 3 projects that use this

  • nicepkgSkill

    team

    根据任务快速组建临时 AI Agent 团队协作。自动从 .claude/agents/ 中选择最合适的成员组队。

    Also in 3 projects that use this

  • sahinSkill

    execute

    Also in 3 projects that use this

  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 5 projects that use this

  • AWSSkill

    retro

    Conduct a retrospective on a coding session.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 3

    Distinctive

    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
  1. Very specific:

    Ask at most three questions, only ones whose answers change the work, with AskUserQuestion so they are cheap to answer.

    line 20
  2. Very specific:

    Use this shape and keep it to a screen.

    line 30
  3. Distinctive:

    Don't restate the stack or conventions.

    line 58
  4. Distinctive:

    Do not write a full intent for a typo, a copy change or a one-file bug: open a short issue (what and why, one line each) and do the work.

    line 79
  5. Distinctive:

    Do not invent acceptance criteria to fill the section.

    line 80
  6. Very specific:

    Do not let the intent grow into a spec.

    line 81

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Write the intent before the work

The issue is where a change's intent lives: it is public, contributors and agents read it, and the pull request closes it. An intent written there before any code is what the diff is judged against.

Argument: an issue number (#14), or a description of the work — then find or open the issue.

1. Find the issue, or open one

  • With a number: gh issue view N --comments. Expect it to be thin; that is normal, not a reason to stop.
  • Without one: gh issue list --search "<words>" --state all — half of the work has been discussed before. Use an open issue that fits; mention a closed one if it is related.
  • Also look for what exists: grep the README, CLAUDE.md, docs/ and the code for the feature's name.

2. Interview, briefly

Ask at most three questions, only ones whose answers change the work, with AskUserQuestion so they are cheap to answer. The ones that usually matter:

  • Who is it for, and what do they do today instead? That is the whole "why".
  • What must not move? Card fields other tools read, the tool API clients rely on, paid calls without consent, English-only product text.
  • Where does it stop? The neighbouring thing it deliberately does not touch.

Decide everything else yourself and write down what you decided. A question the repository can answer is not asked.

3. Write the intent into the issue

Use this shape and keep it to a screen. Create the issue with it (gh issue create --title … --label enhancement|bug|documentation --body …) or replace a thin body (gh issue edit N --body …). Writing to GitHub is outward-facing: ask the user before the first write in a task.

## Want
One paragraph, in terms of what changes for the person using character-forge (a writer in their chat, a contributor), not components.

## Why now
The pain this removes, concretely. If nothing hurts, the work can wait.

## Constraints
What must not move, and what must be reused rather than rebuilt.

## Done when
- [ ] Checkable statements, each true or false by using the tool, the CLI or the site.

## How to check it
Steps anyone can follow to see each "Done when" working: what to ask in the chat, which CLI command to run and what it prints, which page of the site to open. Paid steps: say so and give the cheapest way to check.

## Out of scope
The neighbouring work this deliberately does not do.

## Open questions
Each with the answer assumed for now, so the work is never blocked on one.

Rules that matter more than the shape:

  • Done when items become the acceptance checks and, where they can be, the tests. Three real ones beat ten written to look thorough.
  • Out of scope is not optional: it stops an agent from rebuilding the feature next door.
  • Don't restate the stack or conventions; they are in CLAUDE.md.

4. Open the branch, linked to the issue

git switch main && git pull --ff-only
gh issue develop N --base main --name <type>/N-short-name --checkout

Type: feat, fix, docs, refactor, test, chore. The branch shows in the issue's Development panel, and the pull request will close the issue (Closes #N).

5. Hand off

Continue with the contribute skill from "Develop". Say what comes next and why: a card field → the checklist in CLAUDE.md; site UI → design in Pencil (design/character-forge.pen) first, then code; a new media step → models.json pricing and character_next.

When the work is done or drifts

When the pull request is ready, comment on the issue what was decided along the way and what was deliberately not done (a test not written, a case not handled) — a ticked checklist without comment tells the maintainer nothing. If the work has drifted from the intent, report the drift in a sentence or two and edit the issue body: a stale intent is worse than none.

Do not

  • Do not write a full intent for a typo, a copy change or a one-file bug: open a short issue (what and why, one line each) and do the work. Every change still has an issue.
  • Do not invent acceptance criteria to fill the section.
  • Do not let the intent grow into a spec. When it passes a screen, the rest belongs in docs/<feature>.md.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.