{"id":"1d962c12-0153-4693-bb0f-50a0a60f919a","shortId":"9EFnBv","kind":"skill","title":"prd-builder","tagline":"Production-grade, 10,000+ line PRDs that let AI agents one-shot entire applications. Built on CRHQ.ai.","description":"# PRD Builder Skill\n\n> **Purpose:** Build comprehensive, AI-agent-executable PRDs through a multi-draft, multi-reviewer methodology. The methodology is universal; the output adapts to the project (size, tech stack, scope, type).\n\n---\n\n## When to Use This Skill\n\nTrigger on any of:\n- User asks to write a PRD, product spec, technical spec, or build plan\n- User wants to scope a non-trivial software project or feature\n- User says \"I want an AI agent to build this\" or \"I want to give this to a developer\"\n- User describes a system/feature/product and expects you to formalize it\n\n**Do NOT use for:** trivial bug fixes, single-line changes, or anything where a few sentences in chat suffice.\n\n---\n\n## Core Principles\n\n1. **No ambiguity, ever.** Every contradiction in the PRD becomes a bug in the implementation. The reviewer rounds exist to find and kill these.\n2. **Canonical sources.** Each concern (schema, API, security, naming, etc.) has ONE canonical section. All other sections reference it; they never redefine it.\n3. **Parallel where possible, sequential where dependent.** Drafting and reviewing parallelize naturally. Phases are sequential.\n4. **At least three drafts.** Draft 1 (write) → Reviews → Draft 2 (fixes) → Final Review → Draft 3 (final fixes). Skipping rounds costs more downstream than it saves now.\n5. **Two reviewer perspectives.** A self-review (architect's eye) and a different-perspective review (security/QA/UX/etc.) catch different categories of issues.\n6. **Defaults are decisions.** Every customization question must have a working default so the executing agent never blocks.\n7. **The PRD is also a runbook.** The final section instructs the executor — the PRD is not just *what to build*, it's *how to build it*.\n8. **Right-size everything.** A small feature PRD should not pretend to be a platform spec. The methodology is fixed; the output's scope, depth, and length adapt to the request.\n\n---\n\n## Project Types — Adapt Before Drafting\n\nBefore applying the canonical structure, classify the request. Different types need different emphasis.\n\n| Project Type | What changes |\n|---|---|\n| **New full-stack product** | Full canonical structure applies. All 10+ sections relevant. |\n| **New backend service / API** | Frontend sections are minimal or skipped. API and schema sections expand. |\n| **New frontend / UI app** | Schema may be omitted. Component architecture, state management, design system expand. |\n| **Feature added to existing product** | **MUST analyze the existing codebase first.** PRD references existing patterns, conventions, schema, and adds delta. Many sections become \"follow existing conventions in `<file>`\" rather than restating them. |\n| **Refactor / migration** | Schema diffs, migration plan, rollback plan, testing matrix become primary. |\n| **Internal tool / CLI** | UI sections drop; CLI ergonomics, configuration, distribution take their place. |\n| **Library / SDK** | Public API surface, semver policy, examples, docs expand. Server/UI sections drop. |\n| **Integration / connector** | External service contract, error handling, retries, rate limits become primary. |\n\nIf the request is to extend an existing codebase, **Phase 1 expands** to include a codebase analysis pass before drafting begins. Do not write a PRD that contradicts the code that's already shipped.\n\n---\n\n## The Seven Phases\n\n```\nPhase 1: Requirement Discovery        (parse intent, codebase analysis if applicable, fill gaps)\nPhase 2: Draft 1 Production            (parallel agents, one per section group)\nPhase 3: Dual Review                   (self-review + different-perspective review, parallel)\nPhase 4: Draft 2 Production            (apply ALL fixes, parallel where possible)\nPhase 5: Final Review                  (one reviewer, hunt remaining issues)\nPhase 6: Draft 3 Production            (apply remaining fixes)\nPhase 7: Delivery                       (artifact + download link if available + memory update)\n```\n\n---\n\n## Phase 1 — Requirement Discovery\n\n### Step 1.1: Classify\n\nPick a project type from the table above. This drives everything.\n\n### Step 1.2: For \"feature on existing codebase\" — Analyze First\n\nBefore any drafting, understand the codebase:\n- Tech stack actually in use\n- Folder structure and naming conventions\n- Existing patterns (data access, validation, error handling, auth)\n- Existing schema (tables, columns, FK relationships)\n- Existing API conventions (path style, response shape, auth model)\n- Test setup and conventions\n- Deployment/build scripts\n\nThe PRD must extend these, not contradict them.\n\n### Step 1.3: Parse Intent Against Requirement Dimensions\n\nThese are **prompts, not requirements** — a frontend-only PRD doesn't need a database section. Use as a checklist of what *might* be relevant.\n\n| # | Dimension | Often relevant for |\n|---|-----------|---------------------|\n| 1 | Product/feature type | All |\n| 2 | Core domain entities/objects | Backend, full-stack, data-heavy UIs |\n| 3 | Users & roles | Multi-user products |\n| 4 | Authentication & authorization | Anything with users |\n| 5 | Tech stack hints | All |\n| 6 | External integrations | Anything that talks to other services |\n| 7 | Quality/feature bar | All — but the bar varies by project |\n| 8 | API exposure / public-API posture | Backend, libraries |\n| 9 | Compliance / security | Regulated domains, anything with PII |\n| 10 | Performance / scale targets | Anything user-facing or high-volume |\n| 11 | Deployment model | Productionized work |\n| 12 | Customization scope | Multi-tenant or operator-configurable products |\n| 13 | Execution model (AI vs human, solo vs team) | All — drives Section 10 |\n| 14 | Output / deliverable format | All |\n| 15 | Existing constraints (codebase, infra, brand) | Extension projects |\n\n### Step 1.4: Write a Requirement Summary\n\nCapture in session memory or a working note before drafting. Keep it terse:\n\n```markdown\n## Requirement Summary\n- Type: <one of the project types>\n- Scope: <one paragraph>\n- Key dimensions: <relevant from the list>\n- Existing context: <if extension — what's already there>\n- Open questions: <only ones that genuinely need answers>\n```\n\nIf there are open questions, ask them in **a single batched message**. Never one at a time.\n\n---\n\n## Phase 2 — Draft 1 Production\n\n### Section Structure — Adaptive, Not Fixed\n\nBelow is a **canonical menu** of sections. Use what fits the project. Add sections specific to the project type. Drop or merge sections that don't apply.\n\n| # | Section | When to include |\n|---|---------|------------------|\n| 1 | **Before You Start** (user customization Q&A with defaults) | When the PRD will be executed by another agent or team and customization is expected |\n| 2 | **Project Overview & Vision** | Always |\n| 3 | **Technology Stack & Architecture** | New projects (or extensions where stack matters); skip for pure spec docs |\n| 4 | **Conventions & Best Practices** | Most projects; for extensions, replace with \"Follow existing conventions in `<paths>`\" |\n| 5 | **Data Model / Schema** | When data persistence is involved |\n| 6 | **Core Features / Functional Requirements** | Always |\n| 7 | **API / Interface Design** | When there's a service interface (HTTP, RPC, library, CLI) |\n| 8 | **Security, Privacy, Compliance** | Always — even \"no auth needed\" deserves a sentence |\n| 9 | **Milestones / Execution Plan** | When execution will be sequenced (multi-step build) |\n| 10 | **Executor Instructions** | When an AI agent or external team will execute |\n| + | **Project-specific sections** | Add freely. E.g. \"Design System\", \"ML Model Spec\", \"Migration Plan\", \"Rollout Plan\", \"Telemetry\" |\n\n**The agent has discretion to add, merge, split, or drop sections** based on what the project actually needs. A frontend-only PRD might have: Overview, Design System, Component Architecture, State Management, Routing, Conventions, Milestones, Executor Instructions — and that's the right shape for it.\n\n### Parallel Drafting\n\nSplit the writing across parallel background agents. Group related sections together so each agent has coherent context.\n\nFor each agent:\n- Hand it the section(s) to write with bullet-point requirements\n- Hand it a save location\n- Tell it to be exhaustive within scope: \"every edge case, every validation rule, zero ambiguity *for the things that need to be specified*\"\n- Tell it what's canonical elsewhere so it references rather than redefines\n\nRun the agents in parallel. Wait for completion. Assemble into a single document.\n\n### Per-Section Detail Expectations\n\nCalibrate to the project type. Examples:\n\n- **Schema sections** (when present): every column, type, constraint, index, FK behavior, seed data, migration order\n- **Feature sections**: API endpoints, validation, frontend behavior, side effects, authorization, edge cases — *as relevant*\n- **Component sections** (frontend-heavy): component hierarchy, props, state ownership, interaction patterns\n- **CLI sections**: command tree, flags, exit codes, output formats\n- **Library sections**: public API surface, types, examples, error model\n\nThe principle: **specify what's specifiable, leave room for craft where craft is needed.** Do not force a schema section into a CSS-refactor PRD.\n\n---\n\n## Phase 3 — Dual Review\n\nLaunch two reviewers in parallel with different perspectives. They read the assembled Draft 1 and produce structured findings.\n\n### Reviewer A: Self-Review (Architect's perspective)\n\nLooks for:\n- **Gaps** — what's missing that the project type needs?\n- **Contradictions** — does Section X say something different from Section Y?\n- **Ambiguities** — could a careful executor still misinterpret this?\n- **Sequencing issues** — do milestones depend on things they don't actually have yet?\n- **Structural issues** — broken cross-references, duplicated specs, naming drift\n\n### Reviewer B: Different-Perspective Review\n\nThe right \"different perspective\" depends on the project type. Pick one or compose a hybrid:\n\n| Project type | Suggested perspective |\n|---|---|\n| Full-stack product | Security engineer + QA lead |\n| Backend service | Security + reliability/SRE |\n| Frontend / UI | UX + accessibility + design systems |\n| Library / SDK | API designer + DX (developer experience) |\n| Internal tool | Operator / day-2 maintainer |\n| Data pipeline | Data engineer + cost/perf |\n| Mobile | Platform-conventions expert |\n\nWhichever you pick: instruct the reviewer to **poke holes specifically from that perspective**.\n\n### Required Reviewer Output Format\n\nBoth reviewers produce reports in this shape:\n\n```markdown\n# Review Report\n\n## CRITICAL — Will cause build failures or major rework\n1. <short title> — <where, what, how to fix>\n\n## HIGH — Will cause confusion or significant rework\n...\n\n## MEDIUM / LOW\n...\n\n## Master Contradiction Table\n| # | Topic | Section A says | Section B says | Resolution |\n\n## Priority Action Items (P0 / P1 / P2)\n```\n\nInsist on this format. Vague feedback (\"improve security\") is rejected and re-requested.\n\n---\n\n## Phase 4 — Draft 2 Production\n\nSynthesize ALL findings from both reviewers into a fix list. Then launch parallel agents (typically the same split as Phase 2) with explicit per-section fix instructions.\n\n### The Draft 2 Agent Prompt Pattern\n\n```\nYou are rewriting Section <X> for Draft 2. Read the current draft at <path>.\nApply ALL of the following fixes:\n\n1. <Specific fix — where, what, why>\n2. ...\n\nSave the rewritten section to <new path>.\n```\n\n**Critical:** Make fixes specific and copy-pasteable. Don't say \"fix the password policy\" — say \"harmonize password minimum to N characters in Sections X.Y, X.Z, and N.\" Vague fix instructions create new contradictions.\n\n### Cross-Cutting Standards to Reconcile in Draft 2\n\nThese categories of contradiction tend to appear across nearly every PRD. The *categories* are universal; the *specific values* are project-dependent. Pick one canonical answer per category and enforce it everywhere.\n\n- **Path / endpoint conventions** — pick one style and use it everywhere\n- **Response and error envelope shapes** — define ONCE, reference elsewhere\n- **Pagination model** — pick one approach\n- **Naming conventions** — match what fits the stack and the codebase\n- **Soft vs hard delete** — pick one policy per entity type\n- **Identifier generation** — application-side or DB-side, pick one\n- **Migration / schema-change tooling** — one tool\n- **Validation library** — one library, ideally shareable across boundaries\n- **Configurable vs fixed enums** — be explicit which is which\n\nThese are *categories to check*, not *answers to enforce*. The right answer depends on the codebase, the team, and the constraints.\n\n---\n\n## Phase 5 — Final Review\n\nSingle reviewer, final pass. Goal: catch what the dual review missed.\n\nSpecifically check:\n- Did the Draft 2 fixes actually land?\n- Did any fixes introduce new contradictions?\n- Are all sections internally consistent?\n- Do schema references match the schema definition (if applicable)?\n- Are all environment variables / configuration items documented in one place?\n- Do milestones each have testable exit criteria?\n- Are cross-references between sections still accurate?\n- **Beyond this checklist** — this is the final review and the last chance to catch anything that might still be improved. Look for issues, gaps, ambiguities, or improvements *not* enumerated above. Trust your judgment; flag anything that would make the PRD stronger.\n\nOutput: a concise list of remaining issues — not a massive document. If the doc is solid, the reviewer says so.\n\n---\n\n## Phase 6 — Draft 3 Production\n\nSingle agent applies the final review's remaining fixes. These are usually small (one-line corrections, value swaps, missing items). After this pass:\n- Update the version header to \"Final\" (or whatever finalization marker fits)\n- Save to the working location\n\n---\n\n## Phase 7 — Delivery\n\nThree things should happen, **adapted to whatever environment the skill is running in**:\n\n1. **Show the result to the user.** If the host environment supports inline artifacts/canvases/previews, render the PRD there in editable form. Otherwise, surface it via whatever the environment provides (file path, message attachment, link).\n2. **Make it portable.** If the host environment supports public links/downloads, produce one. Otherwise, ensure the file is saved somewhere the user can retrieve.\n3. **Update working memory.** Capture what was produced, what drafts existed, and key decisions made — so future sessions or audits can reconstruct the reasoning.\n\nThe skill is environment-agnostic. Use the host's facilities for artifacts, downloads, and memory; do not assume specific paths or APIs.\n\n---\n\n## Quality Markers\n\nA finished PRD should pass these checks. Adapt the list — not every marker applies to every project.\n\n- [ ] Each concern has a single canonical section; other sections reference, never redefine\n- [ ] No surviving contradictions from the master contradiction table\n- [ ] All findings from the different-perspective review are addressed or explicitly accepted with rationale\n- [ ] Cross-references between sections are accurate\n- [ ] Every external dependency is named (libraries, services, APIs)\n- [ ] Every configuration item is documented in one place\n- [ ] Every milestone has testable exit criteria (when milestones are present)\n- [ ] Defaults are specified for every customization question (when Q&A is present)\n- [ ] Project-type-relevant cross-cutting concerns are addressed (e.g. accessibility for UIs, rate limits for APIs, observability for services)\n- [ ] No surviving \"TBD\", \"TODO\", or \"we'll figure this out later\"\n- [ ] An executor reading this PRD cold can begin work without immediately needing clarifying questions about the document itself\n- [ ] The PRD respects existing codebase conventions when extending an existing project\n- [ ] The PRD's depth matches the request — not bloated, not skeletal\n\n---\n\n## Anti-Patterns to Avoid\n\n| Anti-pattern | Why it kills the PRD |\n|---|---|\n| Writing one big monolith yourself in serial | Slow, prone to forgetting sections, no parallelism gain |\n| Single-reviewer feedback | Catches one perspective's blind spots only |\n| Skipping Draft 2 | Critical contradictions reach implementation |\n| Vague review feedback (\"improve security\") | Cannot be applied; produces churn |\n| Inline-redefining canonical shapes (response/error/pagination) per section | Guarantees contradiction — define once, reference everywhere |\n| Letting \"TBD\" survive into final draft | Means the executor will guess (badly) |\n| Forcing the canonical 10-section template onto a project that doesn't fit it | Bloat or irrelevance |\n| Specifying tech that ties the PRD to one infrastructure or vendor when the user didn't ask for it | Reduces portability — keep environment-specific bits behind a referenced skill or env var |\n| Asking the user a string of one-at-a-time questions | Disrespects their time; batch them |\n| Drafting before requirement summary is captured | Context drifts mid-draft |\n| Writing a PRD for a feature without first reading the existing codebase | Produces a PRD that contradicts shipped code |\n| Padding to hit a word/line count | Length is a side effect of completeness, not a target |\n\n---\n\n## Reference Workflow Checklist\n\nUse as a live checklist when running the skill. Adapt commands and paths to the host environment.\n\n```\nPhase 1 — Discovery\n  [ ] Classify project type\n  [ ] If extending existing code: analyze the codebase\n  [ ] Parse user prompt against Requirement Dimensions\n  [ ] Write requirement summary to working memory\n  [ ] Ask batched follow-up questions if anything is genuinely unclear\n  [ ] Confirm scope with user before proceeding\n\nPhase 2 — Draft 1\n  [ ] Decide which canonical sections apply; add project-specific ones\n  [ ] Plan section split across parallel drafting agents\n  [ ] Launch all drafting agents in parallel (use whatever async/background mechanism the host provides)\n  [ ] Wait for completions\n  [ ] Assemble Draft 1 into a single document\n\nPhase 3 — Dual Review\n  [ ] Pick a \"different perspective\" appropriate to the project type\n  [ ] Launch self-reviewer + different-perspective reviewer in parallel\n  [ ] Wait for both reports\n  [ ] Read both fully\n\nPhase 4 — Draft 2\n  [ ] Synthesize unified fix list from both reviews\n  [ ] Launch parallel rewrite agents with specific per-section fix instructions\n  [ ] Reconcile cross-cutting categories listed above\n  [ ] Assemble Draft 2\n\nPhase 5 — Final Review\n  [ ] Single reviewer, concise output expected\n  [ ] Identify remaining issues (typically small)\n\nPhase 6 — Draft 3\n  [ ] Apply final fixes\n  [ ] Mark as Final\n  [ ] Save to the working location\n\nPhase 7 — Delivery\n  [ ] Show the result via the host's artifact/preview mechanism (if available)\n  [ ] Produce a download/share link (if the host supports it)\n  [ ] Update working memory with summary of drafts and decisions\n  [ ] Hand off to the user with clear pointers\n```\n\n---\n\n## When to Update This Skill\n\nUpdate when:\n- A pattern emerges across multiple PRDs that should become a default consideration (add to canonical sections menu or reviewer prompts)\n- A category of issue keeps appearing in final review (add it to Phase 4 cross-cutting categories)\n- A user provides post-execution feedback revealing a blind spot (e.g. \"the responsive part wasn't covered\" → add as a project-type consideration)\n- A new project type becomes common (add to the project-types table)\n\nDo NOT update for:\n- Project-specific quirks (those go in the PRD itself, not the skill)\n- One-time formatting preferences\n- Tech-stack-specific opinions (those belong in code-conventions skills, not here)","tags":["prd","builder","crhq","prds","zeropointrepo","agent-skills","agentic-ai","agentic-workflow","ai-development","ai-development-platforms","claude-code","product-requirements"],"capabilities":["skill","source-zeropointrepo","skill-prd-builder","topic-agent-skills","topic-agentic-ai","topic-agentic-workflow","topic-ai-development","topic-ai-development-platforms","topic-claude-code","topic-prd","topic-product-requirements"],"categories":["CRHQ-PRDs"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/ZeroPointRepo/CRHQ-PRDs/prd-builder","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add ZeroPointRepo/CRHQ-PRDs","source_repo":"https://github.com/ZeroPointRepo/CRHQ-PRDs","install_from":"skills.sh"}},"qualityScore":"0.454","qualityRationale":"deterministic score 0.45 from registry signals: · indexed on github topic:agent-skills · 8 github stars · SKILL.md body (19,647 chars)","verified":false,"liveness":"unknown","lastLivenessCheck":null,"agentReviews":{"count":0,"score_avg":null,"cost_usd_avg":null,"success_rate":null,"latency_p50_ms":null,"narrative_summary":null,"summary_updated_at":null},"enrichmentModel":"deterministic:skill-github:v1","enrichmentVersion":1,"enrichedAt":"2026-05-18T19:08:52.354Z","embedding":null,"createdAt":"2026-05-18T13:14:33.137Z","updatedAt":"2026-05-18T19:08:52.354Z","lastSeenAt":"2026-05-18T19:08:52.354Z","tsv":"'-2':1429 '000':8 '1':143,211,488,516,530,589,704,884,922,1310,1476,1579,1957,2449,2493,2529 '1.1':593 '1.2':607 '1.3':669 '1.4':829 '10':7,364,774,814,1036,2318 '11':786 '12':791 '13':802 '14':815 '15':820 '2':167,215,528,553,708,882,947,1525,1547,1557,1567,1585,1633,1786,1991,2274,2491,2567,2595 '3':190,220,539,573,720,952,1294,1899,2015,2535,2613 '4':205,551,727,968,1523,2565,2705 '5':232,562,733,982,1767,2597 '6':255,571,738,991,1897,2611 '7':273,579,747,997,1942,2626 '8':300,757,1011 '9':766,1023 'accept':2113 'access':634,1415,2172 'accur':1834,2122 'across':1115,1641,1734,2507,2675 'action':1503 'actual':623,1081,1362,1788 'ad':398 'adapt':49,328,334,888,1948,2071,2440 'add':415,903,1052,1070,2499,2684,2701,2728,2741 'address':2110,2170 'agent':14,31,98,270,533,940,1042,1066,1118,1125,1131,1186,1540,1558,1902,2510,2514,2578 'agnost':2044 'ai':13,30,97,805,1041 'ai-agent-execut':29 'alreadi':510,860 'also':277 'alway':951,996,1015 'ambigu':145,1163,1344,1859 'analysi':494,522 'analyz':403,613,2458 'anoth':939 'answer':1659,1751,1756 'anti':2234,2239 'anti-pattern':2233,2238 'anyth':133,730,741,771,778,1849,1869,2480 'api':173,370,377,456,646,758,762,998,1225,1261,1420,2061,2130,2178 'app':385 'appear':1640,2697 'appli':338,362,555,575,917,1573,1903,2077,2286,2498,2614 'applic':19,524,1713,1809 'application-sid':1712 'approach':1689 'appropri':2542 'architect':240,1320 'architectur':391,955,1094 'artifact':581,2051 'artifact/preview':2635 'artifacts/canvases/previews':1970 'ask':68,869,2348,2365,2473 'assembl':1192,1308,2527,2593 'assum':2057 'async/background':2519 'attach':1989 'audit':2034 'auth':638,652,1018 'authent':728 'author':729,1232 'avail':585,2638 'avoid':2237 'b':1376,1499 'backend':368,712,764,1408 'background':1117 'bad':2314 'bar':749,753 'base':1076 'batch':874,2380,2474 'becom':152,419,438,476,2680,2739 'begin':498,2200 'behavior':1218,1229 'behind':2358 'belong':2776 'best':970 'beyond':1835 'big':2248 'bit':2357 'blind':2269,2719 'bloat':2230,2329 'block':272 'boundari':1735 'brand':825 'broken':1367 'bug':126,154 'build':27,78,100,293,298,1035,1471 'builder':3,24 'built':20 'bullet':1141 'bullet-point':1140 'calibr':1202 'cannot':2284 'canon':168,179,340,360,894,1176,1658,2086,2292,2317,2496,2686 'captur':834,2019,2387 'care':1347 'case':1158,1234 'catch':250,1775,1848,2265 'categori':252,1635,1646,1661,1747,2590,2693,2709 'caus':1470,1484 'chanc':1846 'chang':131,353,1724 'charact':1612 'chat':139 'check':1749,1782,2070 'checklist':694,1837,2430,2435 'churn':2288 'clarifi':2205 'classifi':342,594,2451 'clear':2663 'cli':442,446,1010,1249 'code':507,1255,2411,2457,2779 'code-convent':2778 'codebas':406,486,493,521,612,620,823,1699,1760,2215,2404,2460 'coher':1127 'cold':2198 'column':642,1213 'command':1251,2441 'common':2740 'complet':1191,2424,2526 'complianc':767,1014 'compon':390,1093,1237,1242 'compos':1393 'comprehens':28 'concern':171,2082,2168 'concis':1878,2602 'configur':448,800,1736,1814,2132 'confirm':2484 'confus':1485 'connector':467 'consider':2683,2734 'consist':1800 'constraint':822,1215,1765 'context':855,1128,2388 'contract':470 'contradict':148,505,666,1334,1492,1624,1637,1795,2095,2099,2276,2298,2409 'convent':412,422,630,647,657,969,980,1098,1439,1668,1691,2216,2780 'copi':1597 'copy-past':1596 'core':141,709,992 'correct':1917 'cost':225 'cost/perf':1435 'could':1345 'count':2417 'cover':2727 'craft':1276,1278 'creat':1622 'crhq.ai':22 'criteria':1826,2144 'critic':1468,1591,2275 'cross':1369,1626,1829,2117,2166,2588,2707 'cross-cut':1625,2165,2587,2706 'cross-refer':1368,1828,2116 'css':1290 'css-refactor':1289 'current':1570 'custom':260,792,927,944,2154 'cut':1627,2167,2589,2708 'data':633,717,983,987,1220,1431,1433 'data-heavi':716 'databas':689 'day':1428 'db':1717 'db-side':1716 'decid':2494 'decis':258,2028,2656 'default':256,266,931,2149,2682 'defin':1681,2299 'definit':1807 'delet':1703 'deliver':817 'deliveri':580,1943,2627 'delta':416 'depend':196,1356,1385,1655,1757,2125 'deploy':787 'deployment/build':658 'depth':325,2225 'describ':112 'deserv':1020 'design':394,1000,1055,1091,1416,1421 'detail':1200 'develop':110,1423 'didn':2346 'diff':431 'differ':246,251,345,348,546,1303,1340,1378,1383,2106,2540,2552 'different-perspect':245,545,1377,2105,2551 'dimens':674,700,853,2466 'discoveri':518,591,2450 'discret':1068 'disrespect':2377 'distribut':449 'doc':461,967,1889 'document':1196,1816,1886,2135,2209,2533 'doesn':685,2325 'domain':710,770 'download':582,2052 'download/share':2641 'downstream':227 'draft':38,197,209,210,214,219,336,497,529,552,572,617,843,883,1111,1309,1524,1556,1566,1571,1632,1785,1898,2024,2273,2308,2382,2392,2492,2509,2513,2528,2566,2594,2612,2654 'drift':1374,2389 'drive':604,812 'drop':445,465,910,1074 'dual':540,1295,1778,2536 'duplic':1371 'dx':1422 'e.g':1054,2171,2721 'edg':1157,1233 'edit':1976 'effect':1231,2422 'elsewher':1177,1684 'emerg':2674 'emphasi':349 'endpoint':1226,1667 'enforc':1663,1753 'engin':1405,1434 'ensur':2005 'entir':18 'entiti':1708 'entities/objects':711 'enum':1739 'enumer':1863 'env':2363 'envelop':1679 'environ':1812,1951,1967,1984,1998,2043,2355,2447 'environment-agnost':2042 'environment-specif':2354 'ergonom':447 'error':471,636,1265,1678 'etc':176 'even':1016 'ever':146 'everi':147,259,1156,1159,1212,1643,2075,2079,2123,2131,2139,2153 'everyth':304,605 'everywher':1665,1675,2302 'exampl':460,1207,1264 'execut':32,269,803,937,1025,1028,1047,2715 'executor':285,1037,1100,1348,2194,2311 'exhaust':1153 'exist':161,400,405,410,421,485,611,631,639,645,821,854,979,2025,2214,2220,2403,2456 'exit':1254,1825,2143 'expand':381,396,462,489 'expect':116,946,1201,2604 'experi':1424 'expert':1440 'explicit':1549,1741,2112 'exposur':759 'extend':483,663,2218,2455 'extens':826,857,959,975 'extern':468,739,1044,2124 'eye':242 'face':781 'facil':2049 'failur':1472 'featur':91,307,397,609,993,1223,2398 'feedback':1513,2264,2281,2716 'figur':2189 'file':1986,2007 'fill':525 'final':217,221,281,563,1768,1772,1841,1905,1930,1933,2307,2598,2615,2619,2699 'find':163,1314,1529,2102 'finish':2065 'first':407,614,2400 'fit':900,1694,1935,2327 'fix':127,216,222,320,557,577,890,1481,1535,1553,1578,1581,1593,1602,1620,1738,1787,1792,1909,2570,2584,2616 'fk':643,1217 'flag':1253,1868 'folder':626 'follow':420,978,1577,2476 'follow-up':2475 'forc':1283,2315 'forget':2256 'form':1977 'formal':119 'format':818,1257,1457,1511,2768 'freeli':1053 'frontend':371,383,682,1085,1228,1240,1412 'frontend-heavi':1239 'frontend-on':681,1084 'full':356,359,714,1401 'full-stack':355,713,1400 'fulli':2563 'function':994 'futur':2031 'gain':2260 'gap':526,1325,1858 'generat':1711 'genuin':2482 'give':106 'go':2757 'goal':1774 'grade':6 'group':537,1119 'guarante':2297 'guess':2313 'hand':1132,1144,2657 'handl':472,637 'happen':1947 'hard':1702 'harmon':1607 'header':1928 'heavi':718,1241 'hierarchi':1243 'high':784,1482 'high-volum':783 'hint':736 'hit':2414 'hole':1449 'host':1966,1997,2047,2446,2522,2633,2645 'http':1007 'human':807 'hunt':567 'hybrid':1395 'ideal':1732 'identifi':1710,2605 'immedi':2203 'implement':157,2278 'improv':1514,1854,1861,2282 'includ':491,921 'index':1216 'infra':824 'infrastructur':2340 'inlin':1969,2290 'inline-redefin':2289 'insist':1508 'instruct':283,1038,1101,1444,1554,1621,2585 'integr':466,740 'intent':520,671 'interact':1247 'interfac':999,1006 'intern':440,1425,1799 'introduc':1793 'involv':990 'irrelev':2331 'issu':254,569,1353,1366,1857,1882,2607,2695 'item':1504,1815,1921,2133 'judgment':1867 'keep':844,2353,2696 'key':852,2027 'kill':165,2243 'land':1789 'last':1845 'later':2192 'launch':1297,1538,2511,2547,2575 'lead':1407 'least':207 'leav':1273 'length':327,2418 'let':12,2303 'librari':453,765,1009,1258,1418,1729,1731,2128 'limit':475,2176 'line':9,130,1916 'link':583,1990,2642 'links/downloads':2001 'list':1536,1879,2073,2571,2591 'live':2434 'll':2188 'locat':1148,1940,2624 'look':1323,1855 'low':1490 'made':2029 'maintain':1430 'major':1474 'make':1592,1872,1992 'manag':393,1096 'mani':417 'mark':2617 'markdown':847,1465 'marker':1934,2063,2076 'massiv':1885 'master':1491,2098 'match':1692,1804,2226 'matrix':437 'matter':962 'may':387 'mean':2309 'mechan':2520,2636 'medium':1489 'memori':586,837,2018,2054,2472,2650 'menu':895,2688 'merg':912,1071 'messag':875,1988 'methodolog':42,44,318 'mid':2391 'mid-draft':2390 'might':697,1088,1851 'migrat':429,432,1060,1221,1721 'mileston':1024,1099,1355,1821,2140,2146 'minim':374 'minimum':1609 'misinterpret':1350 'miss':1328,1780,1920 'ml':1057 'mobil':1436 'model':653,788,804,984,1058,1266,1686 'monolith':2249 'multi':37,40,724,795,1033 'multi-draft':36 'multi-review':39 'multi-step':1032 'multi-ten':794 'multi-us':723 'multipl':2676 'must':262,402,662 'n':1611,1618 'name':175,629,1373,1690,2127 'natur':201 'near':1642 'need':347,687,1019,1082,1168,1280,1333,2204 'never':187,271,876,2091 'new':354,367,382,956,1623,1794,2736 'non':86 'non-trivi':85 'note':841 'observ':2179 'often':701 'omit':389 'one':16,178,534,565,877,1391,1657,1670,1688,1705,1720,1726,1730,1818,1915,2003,2137,2247,2266,2339,2372,2503,2766 'one-at-a-tim':2371 'one-lin':1914 'one-shot':15 'one-tim':2765 'onto':2321 'open':862,867 'oper':799,1427 'operator-configur':798 'opinion':2774 'order':1222 'otherwis':1978,2004 'output':48,322,816,1256,1456,1876,2603 'overview':949,1090 'ownership':1246 'p0':1505 'p1':1506 'p2':1507 'pad':2412 'pagin':1685 'parallel':191,200,532,549,558,1110,1116,1188,1301,1539,2259,2508,2516,2556,2576 'pars':519,670,2461 'part':2724 'pass':495,1773,1924,2068 'password':1604,1608 'pasteabl':1598 'path':648,1666,1987,2059,2443 'pattern':411,632,1248,1560,2235,2240,2673 'per':535,1198,1551,1660,1707,2295,2582 'per-sect':1197,1550,2581 'perform':775 'persist':988 'perspect':235,247,547,1304,1322,1379,1384,1399,1453,2107,2267,2541,2553 'phase':202,487,514,515,527,538,550,561,570,578,588,881,1293,1522,1546,1766,1896,1941,2448,2490,2534,2564,2596,2610,2625,2704 'pick':595,1390,1443,1656,1669,1687,1704,1719,2538 'pii':773 'pipelin':1432 'place':452,1819,2138 'plan':79,433,435,1026,1061,1063,2504 'platform':315,1438 'platform-convent':1437 'point':1142 'pointer':2664 'poke':1448 'polici':459,1605,1706 'portabl':1994,2352 'possibl':193,560 'post':2714 'post-execut':2713 'postur':763 'practic':971 'prd':2,23,72,151,275,287,308,408,503,661,684,934,1087,1292,1644,1874,1973,2066,2197,2212,2223,2245,2337,2395,2407,2760 'prd-builder':1 'prds':10,33,2677 'prefer':2769 'present':1211,2148,2160 'pretend':311 'primari':439,477 'principl':142,1268 'prioriti':1502 'privaci':1013 'proceed':2489 'produc':1312,1460,2002,2022,2287,2405,2639 'product':5,73,358,401,531,554,574,726,801,885,1403,1526,1900 'product/feature':705 'production':789 'production-grad':4 'project':52,89,332,350,597,756,827,902,908,948,957,973,1049,1080,1205,1331,1388,1396,1654,2080,2162,2221,2323,2452,2501,2545,2732,2737,2745,2753 'project-depend':1653 'project-specif':1048,2500,2752 'project-typ':2731,2744 'project-type-relev':2161 'prompt':677,1559,2463,2691 'prone':2254 'prop':1244 'provid':1985,2523,2712 'public':455,761,1260,2000 'public-api':760 'pure':965 'purpos':26 'q':928,2157 'qa':1406 'qualiti':2062 'quality/feature':748 'question':261,863,868,2155,2206,2376,2478 'quirk':2755 'rate':474,2175 'rather':424,1181 'rational':2115 're':1520 're-request':1519 'reach':2277 'read':1306,1568,2195,2401,2561 'reason':2038 'reconcil':1630,2586 'reconstruct':2036 'redefin':188,1183,2092,2291 'reduc':2351 'refactor':428,1291 'refer':184,409,1180,1370,1683,1803,1830,2090,2118,2301,2428 'referenc':2360 'regul':769 'reject':1517 'relat':1120 'relationship':644 'relev':366,699,702,1236,2164 'reliability/sre':1411 'remain':568,576,1881,1908,2606 'render':1971 'replac':976 'report':1461,1467,2560 'request':331,344,480,1521,2228 'requir':517,590,673,679,832,848,995,1143,1454,2384,2465,2468 'resolut':1501 'respect':2213 'respons':650,1676,2723 'response/error/pagination':2294 'restat':426 'result':1960,2630 'retri':473 'retriev':2014 'reveal':2717 'review':41,159,199,213,218,234,239,248,541,544,548,564,566,1296,1299,1315,1319,1375,1380,1446,1455,1459,1466,1532,1769,1771,1779,1842,1893,1906,2108,2263,2280,2537,2550,2554,2574,2599,2601,2690,2700 'rework':1475,1488 'rewrit':1563,2577 'rewritten':1588 'right':302,1106,1382,1755 'right-siz':301 'role':722 'rollback':434 'rollout':1062 'room':1274 'round':160,224 'rout':1097 'rpc':1008 'rule':1161 'run':1184,1955,2437 'runbook':279 'save':230,1147,1586,1936,2009,2620 'say':93,1338,1497,1500,1601,1606,1894 'scale':776 'schema':172,379,386,413,430,640,985,1208,1285,1723,1802,1806 'schema-chang':1722 'scope':56,83,324,793,851,1155,2485 'script':659 'sdk':454,1419 'section':180,183,282,365,372,380,418,444,464,536,690,813,886,897,904,913,918,1051,1075,1121,1135,1199,1209,1224,1238,1250,1259,1286,1336,1342,1495,1498,1552,1564,1589,1614,1798,1832,2087,2089,2120,2257,2296,2319,2497,2505,2583,2687 'secur':174,768,1012,1404,1410,1515,2283 'security/qa/ux/etc':249 'seed':1219 'self':238,543,1318,2549 'self-review':237,542,1317,2548 'semver':458 'sentenc':137,1022 'sequenc':1031,1352 'sequenti':194,204 'serial':2252 'server/ui':463 'servic':369,469,746,1005,1409,2129,2181 'session':836,2032 'setup':655 'seven':513 'shape':651,1107,1464,1680,2293 'shareabl':1733 'ship':511,2410 'shot':17 'show':1958,2628 'side':1230,1714,1718,2421 'signific':1487 'singl':129,873,1195,1770,1901,2085,2262,2532,2600 'single-lin':128 'single-review':2261 'size':53,303 'skelet':2232 'skill':25,62,1953,2040,2361,2439,2669,2764,2781 'skill-prd-builder' 'skip':223,376,963,2272 'slow':2253 'small':306,1913,2609 'soft':1700 'softwar':88 'solid':1891 'solo':808 'someth':1339 'somewher':2010 'sourc':169 'source-zeropointrepo' 'spec':74,76,316,966,1059,1372 'specif':905,1050,1450,1580,1594,1650,1781,2058,2356,2502,2580,2754,2773 'specifi':1171,1269,1272,2151,2332 'split':1072,1112,1544,2506 'spot':2270,2720 'stack':55,357,622,715,735,954,961,1402,1696,2772 'standard':1628 'start':925 'state':392,1095,1245 'step':592,606,668,828,1034 'still':1349,1833,1852 'string':2369 'stronger':1875 'structur':341,361,627,887,1313,1365 'style':649,1671 'suffic':140 'suggest':1398 'summari':833,849,2385,2469,2652 'support':1968,1999,2646 'surfac':457,1262,1979 'surviv':2094,2183,2305 'swap':1919 'synthes':1527,2568 'system':395,1056,1092,1417 'system/feature/product':114 'tabl':601,641,1493,2100,2747 'take':450 'talk':743 'target':777,2427 'tbd':2184,2304 'team':810,942,1045,1762 'tech':54,621,734,2333,2771 'tech-stack-specif':2770 'technic':75 'technolog':953 'telemetri':1064 'tell':1149,1172 'templat':2320 'tenant':796 'tend':1638 'ters':846 'test':436,654 'testabl':1824,2142 'thing':1166,1358,1945 'three':208,1944 'tie':2335 'time':880,2375,2379,2767 'todo':2185 'togeth':1122 'tool':441,1426,1725,1727 'topic':1494 'topic-agent-skills' 'topic-agentic-ai' 'topic-agentic-workflow' 'topic-ai-development' 'topic-ai-development-platforms' 'topic-claude-code' 'topic-prd' 'topic-product-requirements' 'tree':1252 'trigger':63 'trivial':87,125 'trust':1865 'two':233,1298 'type':57,333,346,351,598,706,850,909,1206,1214,1263,1332,1389,1397,1709,2163,2453,2546,2733,2738,2746 'typic':1541,2608 'ui':384,443,719,1413,2174 'unclear':2483 'understand':618 'unifi':2569 'univers':46,1648 'updat':587,1925,2016,2648,2667,2670,2750 'use':60,123,625,691,898,1673,2045,2431,2517 'user':67,80,92,111,721,725,732,780,926,1963,2012,2345,2367,2462,2487,2661,2711 'user-fac':779 'usual':1912 'ux':1414 'vagu':1512,1619,2279 'valid':635,1160,1227,1728 'valu':1651,1918 'var':2364 'vari':754 'variabl':1813 'vendor':2342 'version':1927 'via':1981,2631 'vision':950 'volum':785 'vs':806,809,1701,1737 'wait':1189,2524,2557 'want':81,95,104 'wasn':2725 'whatev':1932,1950,1982,2518 'whichev':1441 'within':1154 'without':2202,2399 'word/line':2416 'work':265,790,840,1939,2017,2201,2471,2623,2649 'workflow':2429 'would':1871 'write':70,212,501,830,1114,1138,2246,2393,2467 'x':1337 'x.y':1615 'x.z':1616 'y':1343 'yet':1364 'zero':1162","prices":[{"id":"80b16e54-cc8c-4225-8f8b-9fee9e006226","listingId":"1d962c12-0153-4693-bb0f-50a0a60f919a","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"ZeroPointRepo","category":"CRHQ-PRDs","install_from":"skills.sh"},"createdAt":"2026-05-18T13:14:33.137Z"}],"sources":[{"listingId":"1d962c12-0153-4693-bb0f-50a0a60f919a","source":"github","sourceId":"ZeroPointRepo/CRHQ-PRDs/prd-builder","sourceUrl":"https://github.com/ZeroPointRepo/CRHQ-PRDs/tree/main/skills/prd-builder","isPrimary":false,"firstSeenAt":"2026-05-18T13:14:33.137Z","lastSeenAt":"2026-05-18T19:08:52.354Z"}],"details":{"listingId":"1d962c12-0153-4693-bb0f-50a0a60f919a","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"ZeroPointRepo","slug":"prd-builder","github":{"repo":"ZeroPointRepo/CRHQ-PRDs","stars":8,"topics":["agent-skills","agentic-ai","agentic-workflow","ai-development","ai-development-platforms","claude-code","prd","product-requirement-document","product-requirements","product-requirements-prompt"],"license":null,"html_url":"https://github.com/ZeroPointRepo/CRHQ-PRDs","pushed_at":"2026-05-10T05:09:04Z","description":"Production-grade, 10,000+ line PRDs that let AI agents one-shot entire applications. Built on CRHQ.ai.","skill_md_sha":"b60c40ccec8c623d754331af5b484741543ef388","skill_md_path":"skills/prd-builder/SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/ZeroPointRepo/CRHQ-PRDs/tree/main/skills/prd-builder"},"layout":"multi","source":"github","category":"CRHQ-PRDs","frontmatter":{},"skills_sh_url":"https://skills.sh/ZeroPointRepo/CRHQ-PRDs/prd-builder"},"updatedAt":"2026-05-18T19:08:52.354Z"}}