agentleFS
tornikebolokadze1-cybertornikebolokadze1-cyber/agent-foundation-kit/.claude/skills/intent/SKILL.md

intent

Skill6 starsChanged 27 days agoMITView on GitHub

Structure or update the owner's intent in intent.md — vision, measurable goals, non-goals, constraints, working principles, open questions. Use when the owner describes what the product or the way of working should be, answers interview questions, or wants to change the goal.

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

  • sahinSkill

    execute

    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

  • 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

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 1

    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:

    Reopen if needed: If intent.md is accepted and must change: first an edit that ONLY sets status: draft (the guard allows exactly that), …

    line 14
  2. Very specific:

    The idea first (ADR-0011): If §1 still holds only the template's _…_ line and the person has just described what they want to build, write …

    line 16
  3. Distinctive:

    How to ask: Ask so that a plain „კი" agrees with your recommendation, and record such a yes as

    line 27
  4. Very specific:

    Builder check: When §7 holds no question any more, run the builder check (builder-check.md) on intent.md before you offer /approve intent, …

    line 34

“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

/intent — the owner's intent, structured

intent.md is the root of the chain intent → spec → plan. You write drafts; only the owner accepts (/approve intent). You never set status: accepted.

  1. Read intent.md, docs/domain/*.md, the "ახლა" block of progress.md, and bash .claude/hooks/py.sh tools/context_graph.py query intent.
  2. Reopen if needed. If intent.md is accepted and must change: first an edit that ONLY sets status: draft (the guard allows exactly that), then the content edits.
  3. The idea first (ADR-0011). If §1 still holds only the template's _…_ line and the person has just described what they want to build, write it into §1 at once — their own words, spelling lightly corrected, source მფლობელი with the date — before the first question. Never ask for setup, a README step or an approval of the engine's own ADRs. Write the project's type into the frontmatter — kind: website / game / agent / automation / other (the table in docs/guides/other.md decides a mixed idea) — and take the interview's questions from that type's guide, docs/guides/<kind>.md §1.
  4. Interview when information is missing — ONE question per message (never two), in Georgian, with 2–3 concrete options and your recommendation. Once kind is set, the guide's §1 questions replace the template's list in §7: rewrite §7 as the guide's questions the idea has not answered yet, ask every one of them (none is skipped as obvious), and when one is answered, write the answer where it belongs and delete the question from §7. A question about the stack (what to build it with) waits for stage 3. How to ask. Ask so that a plain „კი" agrees with your recommendation, and record such a yes as მფლობელი (დაეთანხმა რეკომენდაციას) with the date — the choice is the owner's, the words are yours. „როგორც შენ გინდა" or „არ ვიცი" is not a choice: say in one sentence what your recommendation decides and ask again. When the person is stuck, ask about a concrete moment („ბოლოს როდის დაგჭირდა…?") or offer a guess for them to correct. When they answer in their own words, drop the options and follow their words. A new wish that comes up mid-interview goes at once into §4 as „მოგვიანებით" with its date, and the interview goes on. The interview is done when nobody will ever have to ask what the person really meant. Builder check. When §7 holds no question any more, run the builder check (builder-check.md) on intent.md before you offer /approve intent, and sort what it asks into three piles: answered already (the conversation or a document holds the answer — write it in); minor (a sensible builder could settle it without changing what gets built — decide it yourself and mark it ჰიპოთეზა); theirs (only the person can answer it, and the answer changes what gets built). Ask the person's pile in one message, most important first, at most three questions, each with its options and your recommendation — the one exception to one question per message, because it is the last round. Write the person's pile into §7 as its open questions, so the stage stays at the interview until they are answered, and write the check into the history (step 6) as - <version> — <date> — მშენებლის შემოწმება: <N> კითხვა — <a> უკვე ნაპასუხები, <b> თავად გადავწყვიტე (ჰიპოთეზა), <c> ადამიანს ვკითხე: until that line is there, the stage block names only the check and never /approve intent. When you offer /approve intent, name in a sentence or two the calls you made yourself, so the person can check them.
  5. Keep the fixed structure (§1 ხედვა … §8 წყაროები + ცვლილებების ისტორია). Goals keep stable IDs (G1, G2…) — never renumber; retire one with strike-through and the reason. Every goal has "how we measure". Every item names its source: მფლობელი (the owner said it — with the date) or ჰიპოთეზა (inferred — needs confirmation). Never present a hypothesis as the owner's words.
  6. Version — bump version (0.1 → 0.2), set updated, add one line to the history.
  7. Check — bash .claude/hooks/py.sh tools/context_graph.py check intent.md; fix every error.
  8. Tell the owner in 2–3 Georgian sentences what changed and what to review, and the next step from bash .claude/hooks/py.sh tools/context_graph.py stage; the decision is /approve intent.

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.