agentleFS
igovetigovet/cortex-orch-acp/skills/requirements-intake/SKILL.md

requirements-intake

Skill1 starsChanged 59 days agoMITView on GitHub

Clarify materially incomplete requests of any kind before planning, using existing project evidence first and bounded structured questions only for remaining decisions.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use before planning a new request whose goal lacks information that would materially change the result.

    line 8
  2. Distinctive:

    Do not interrogate the user when the request and project evidence suffice.

    line 10
  3. Distinctive:

    Read the request, retained answers, project instructions and relevant product docs.

    line 14
  4. Distinctive:

    Never ask what docs, code or a retained answer already establish.

    line 24
  5. Distinctive:

    Read linked package references through Cortex read_skill using question-banks or agreed-spec (also accepted in references/<id>.md form), …

    line 26
  6. Distinctive:

    Ask at most two rounds, normally one, with at most five material questions per round (or fewer if the live schema requires it).

    line 43
  7. Distinctive:

    If a critical decision is still unresolved after the bound, do not guess or implement it.

    line 46

≈ “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.

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

Requirements intake

Use before planning a new request whose goal lacks information that would materially change the result. Apply to every request kind, not just design. Do not interrogate the user when the request and project evidence suffice.

Check sufficiency first

Read the request, retained answers, project instructions and relevant product docs. Use bounded code-index orientation for existing behavior and constraints; do not turn intake into a coordinator implementation investigation. If locating facts needs substantial investigation, propose that bounded investigation rather than asking the user to reconstruct the repository.

Separate established facts, decisions the user must make, and safe implementation details delegated to a specialist. Check goal, affected users, scope and non-goals, observable acceptance, constraints, risks, and verification. Missing information matters only if different answers would change scope, behavior, risk or acceptance. Never ask what docs, code or a retained answer already establish.

Read linked package references through Cortex read_skill using question-banks or agreed-spec (also accepted in references/<id>.md form), not project_read on the target project.

Ask only material gaps

  1. Select only relevant prompts from the question banks. For feature/product ask outcome and users; for bug/incident ask expected versus actual behavior and reproduction; for refactor ask invariants and permitted changes; for UI/UX ask audience, platforms, screens and visual direction; for documentation ask audience and authoritative facts; for infra/ops ask target, authority and rollback; for investigation ask the decision and evidence needed.
  2. Group related questions into one request through the advertised coordinator question tool. Read its live schema; do not use a native engine question tool or prose-only questions. Use the runtime-selected user question language. Offer meaningful, distinct options grounded in the project, explain tradeoffs, and preserve custom free text. Do not present invented requirements as defaults.
  3. Ask at most two rounds, normally one, with at most five material questions per round (or fewer if the live schema requires it). End the turn immediately after asking. On resumption, incorporate real answers and ask only remaining gaps.
  4. If a critical decision is still unresolved after the bound, do not guess or implement it. Offer a bounded discovery-only scope with no product commitment, or report the concrete missing decision and stop the implementation proposal. A user explicitly delegating a choice is a decision; label the resulting specialist proposal rather than inventing functionality.

Carry the agreement forward

Record the agreed specification in plan objectives: goal, users, scope/non-goals, acceptance examples, constraints, risks, verification expectations and unresolved items. Preserve exact source references for consequential decisions. Use the spec outline as a completeness check, not a mandate to create a document for every small fix.

For durable product or design intake, propose a writing assignment to record the agreed facts in the target project's documentation before dependent work. Read-only designers return specifications in reports; writers persist them. Do not modify lifecycle, permissions or question contracts, promise a tool capability not observed in this session, or treat answered intake as runtime acceptance.

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.