AGENT SKILL

requirement-enrichment

Phrases the pipeline's clarification questions as plain-language two-layer question objects. Use when the label step rated the request vague or okay, unlocked a question budget, and the touched area is grounded.

License 许可
MIT
所属仓库
Eliyce/paqad-ai
最近更新
2026-11-13
Stars
★ 8
Forks
0

01 前言 Front Matter

name
requirement-enrichment
metadata
{ "cacheable": "false", "model_tier": "fast", "paqad-meta": "skill.meta.yaml", "output_format": "json" }
description
Phrases the pipeline's clarification questions as plain-language two-layer question objects. Use when the label step rated the request vague or okay, unlocked a question budget, and the touched area is grounded.

02 资源文件 6

03 SKILL.md 正文

What It Does

Turns the open decisions an incomplete request leaves behind into the pipeline's S2 clarification batch: a set of FR-4 two-layer question objects, each phrased in the project's own words so a non-technical owner can answer it. This is the S2 question-phrasing step of the spec pipeline, it phrases, it does not judge clarity and does not re-run the plain-language check.

Use This When

When the spec pipeline's label step rated the request vague or okay and unlocked a question budget, and the touched area has been grounded (S0). It is for phrasing genuine ambiguities as answerable questions, not for restating a request that is already clear.

Inputs

  • Read the S0 grounding terms and references for the touched area: they are the first-priority vocabulary.
  • Read the raw request, tracker notes, and any linked issue or spec: the user's own wording is the second-priority vocabulary.
  • Read references/enrichment-checklist.md and walk assets/operational-checklist.txt to find the dimensions the request leaves undecided.

Procedure

  1. Walk assets/operational-checklist.txt (permissions, auditability, rollback, data-retention, i18n, a11y, observability, docs, feature-flags) and references/enrichment-checklist.md; for every dimension the request touches but leaves undecided, hold a candidate question.
  2. Keep only genuine ambiguities, within the question budget the label step unlocked. Drop anything the grounding or the prompt already answers.
  3. Phrase each question as a two-layer object. business_text and every options[] entry draw vocabulary in priority order: (1) the S0 grounding terms, (2) the user's own prompt, (3) plain English. Given a documented term ("archived invoices"), use it, never a from-the-model synonym ("soft-deleted records").
  4. Phrase options[] as OUTCOMES, never mechanisms: "keep trying quietly for an hour, then notify someone", never "exponential backoff". Give at least two options.
  5. Where current behaviour is documented, cite it in the question ("Today, exports include hidden columns: keep that, or leave them out?").
  6. Write one plain sentence of why_it_matters. Set grounded_in to the doc/glossary ref the wording came from, or null when it cannot be tied to project evidence. Put any internal mechanism note in technical_note (never shown to the user).
  7. Emit the batch per assets/output.template.md; validate the shape with scripts/lint-output.sh (exit 0). Hand the file to paqad-ai spec pipeline record questions <file>, which runs the ledger auto-answer pre-step and persists the surviving batch.
  8. These questions are not asked on their own. They join the change's one question round at the spec questions step, together with every pending packet paqad-ai spec pipeline round lists for this change (expert picks, the deferred module attribution, the fix proof method, the visual-evidence readiness pause). The owner is asked once, up to 4 questions per call, back to back.

Output Contract

  • A JSON object matching assets/output.template.md: { "questions": [ ... ] }, where each question has business_text (plain string), why_it_matters (one sentence), options[] (>=2 outcome strings), grounded_in (a ref string or null), and an optional internal technical_note.
  • questions[] is the only hard-required field, so the raw batch validates against the pipeline's questions step even before auto-answer enrichment.
  • Do NOT re-implement the plain-language check here: the pipeline's checkPlainLanguage guards the wording (FR-4.3). Phrase for it.
  • Output must pass scripts/lint-output.sh (exit 0).

Escalate / Stop Conditions

  • Ask when business rules, permissions, or data-retention expectations are absent but clearly required: phrase the ask as one of these questions rather than guessing.
  • Warn when the request conflicts with canonical docs or previously approved behavior.
  • Do not guess regulatory, billing, or irreversible-data requirements; surface them as questions with grounded_in set to the governing doc.

Resources

  • references/enrichment-checklist.md
  • scripts/lint-output.sh
  • assets/output.template.md
  • assets/operational-checklist.txt
  • runtime/capabilities/coding/checklists/edge-cases-coding.md
  • agents/openai.yaml

04 同仓库更多 Skills

Eliyce
Eliyce/paqad-ai

sample-skill

Sample bundled skill fixture

★ 8
Eliyce
Eliyce/paqad-ai

processing-invoices

Turns a supplier invoice into one ledger line. Use when a supplier invoice arrives and needs booking.

★ 8
Eliyce
Eliyce/paqad-ai

processing-invoices

Turns a supplier invoice into one ledger line. Use when a supplier invoice arrives and needs booking.

★ 8
Eliyce
Eliyce/paqad-ai

understand-idea

Chat and Claude Code: Step 1 of the discover journey: clarifies the goal, who benefits, success and constraints into the idea block. Use when the journey directs it to step 1, not directly.

★ 8
Eliyce
Eliyce/paqad-ai

size-work

Chat and Claude Code: Suggest a size for a ticket from comparable finished work, as a labelled suggestion the team still owns. Use it when someone asks "how big is this?", "roughly how much work is ABC-12?", or wants a backlog item sized. It finds the team's sizing scale at first use, picks two to five of the closest finished tickets with links, and gives a size or range with the drivers behind it and a confidence of low, medium or high, lowered and explained when the area's documents are template text. Every result is labelled a suggested size, never a date or a commitment. It writes nothing into the size field unless the user says "set it", and it refuses to overwrite an existing size unless the user asks to replace it. Produces the sizing block of the record.

★ 8
Eliyce
Eliyce/paqad-ai

refine-ticket

Chat and Claude Code: Improve one or more existing tracker tickets without losing the author's intent. Use it when someone wants a ticket read, diagnosed and tidied up ("this ticket is vague", "fix up ABC-12", "split this, it's too big"). It reads each ticket in full and snapshots it with its last-updated time, says what is wrong with evidence, and proposes a field-by-field before and after change set limited to summary, description, labels, parent, links and new child tickets on a split. It writes nothing until an explicit "apply it" bound to that exact change set, re-checks the ticket for conflicting edits just before writing, never replaces rich descriptions with plain text, never deletes or closes a ticket, and reads each ticket back after applying. Produces the existing and changes blocks of the record.

★ 8