跳到正文
cccskills
免费在 GitHub 公开

requirements

Turns a feature request into a full implementation-plan requirements document matching docs/plan/_TEMPLATE.md. Use when the user asks for requirements, a spec, an implementation plan, or wants to flesh out a feature before building. Also use for /requirements or "write requirements for X".

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md7.6 KB
  • references/checklist.md2.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Requirements

Turn a user request into a complete Lextures implementation plan — the same structure and rigor as docs/plan/.

Output

Produce a markdown document that follows docs/plan/_TEMPLATE.md exactly. Every section must be filled; no … or TBD placeholders unless listed explicitly in Open Questions (§18).

Default: paste the full document in chat. If the user asks to save it, write to docs/plan/{section-folder}/{id}-{kebab-slug}.md per docs/plan/README.md conventions.

Workflow

Copy this checklist and track progress:

Requirements progress:
- [ ] Step 1: Parse the request
- [ ] Step 2: Explore codebase & existing plans
- [ ] Step 3: Draft metadata
- [ ] Step 4: Write all 19 sections
- [ ] Step 5: Validate against checklist

Step 1. Parse the request

Extract from the user's message:

  • Feature name — short, descriptive title
  • Scope — what they want; what they explicitly excluded
  • Audience — K12 / HE / SL (self-learner) / platform-wide
  • Urgency signals — blocker vs enhancement, pilot vs GA

If critical details are missing (audience, scope boundary, or whether this is net-new vs extending existing behavior), ask one focused question before proceeding. Otherwise state your interpretation briefly and continue.

Step 2. Explore codebase & existing plans

Before writing requirements, ground the spec in what exists:

  1. Read docs/plan/_TEMPLATE.md for the canonical section list.
  2. Search docs/plan/ for related plans — read 1–2 closest matches for tone, cross-references, and dependency patterns.
  3. Search docs/completed/ if the feature may partially exist.
  4. Explore the codebase — Glob/Grep/Read for:
    • Existing routes, services, DB tables, UI pages that overlap
    • Patterns to extend (auth scopes, job queue, feature flags, migrations)
    • File paths to cite in §19 References

Record findings: what is MISSING, PARTIAL, or THIN today (for metadata Status).

Step 3. Draft metadata

Fill the metadata table. Propose values when the user did not supply them:

FieldGuidance
Feature ID{section}.{number} — match an existing docs/plan/ section when possible; otherwise propose next number in the closest section
SectionTop-level plan folder name (e.g. Platform, Performance & Operability)
SeverityBLOCKER / MAJOR / MINOR per docs/plan/README.md legend
MarketsK12 / HE / SL — comma-separated
Status (today)MISSING / PARTIAL / THIN based on codebase exploration
Estimated effortXS / S / M / L / XL
Depends on / UnblocksFeature IDs from related plans; use relative links

Title line format:

# {Feature ID} — {Feature Name}

> Implementation plan. Source: user request / {related doc if any}.

Step 4. Write all 19 sections

Follow the template section-by-section. Quality bar matches completed plans like 17.14-feature-flags.md and 19.3-ai-assisted-grading.md.

§1 Problem Statement — 2–4 sentences. Gap today, who is hurt, business outcome.

§2 Goals — 3–5 outcome bullets. Verifiable, not implementation steps.

§3 Non-Goals — Explicit out-of-scope items. Reference other Feature IDs when deferring work.

§4 Personas & User Stories — As a {role}, I want … so that …. Cover student, instructor, admin, and self-learner where relevant.

§5 Functional Requirements — Numbered FR-N. RFC 2119: MUST / SHOULD / MAY. Testable, not vague.

§6 Non-Functional Requirements — Cover applicable sub-bullets from the template (performance, security, privacy, a11y, scalability, reliability, observability, maintainability, i18n, backward compatibility). Omit sub-bullets that genuinely do not apply; say "N/A" with one-line rationale.

§7 Acceptance Criteria — Numbered AC-N. Given/When/Then. Each AC should map to at least one test.

§8 Data Model — SQL or schema description. Migration naming: server/migrations/NNN_*.sql. Indexes, constraints, backfill strategy.

§9 API Surface — Table or list: method, path, auth scope, request/response shapes. Note rate limits and OpenAPI updates.

§10 UI / UX — Pages, components, flows, empty/loading/error states, responsive behavior, a11y notes, i18n keys.

§11 AI / ML Considerations — Skip entirely (write "N/A — not AI-touching.") only when the feature has zero AI/LLM/ML involvement. Otherwise: models, prompts, eval metrics, fallback, PII, cost budget.

§12 Integration Points — External services and internal modules with file paths.

§13 Dependencies & Sequencing — Must ship after/before; shared infra (Redis, job queue, object storage, etc.).

§14 Risks & Mitigations — Table: Risk | Likelihood | Impact | Mitigation.

§15 Rollout Plan — Feature flag name, phases, pilot cohort, GA criteria, rollback path.

§16 Test Plan — Unit, integration, e2e (Playwright), security, accessibility (axe), performance/load, manual QA.

§17 Documentation & Training — User docs, admin docs, API reference, runbooks.

§18 Open Questions — Numbered decisions still needing owners. Only place for unresolved TBDs.

§19 References — Existing code paths (server/internal/..., clients/web/src/...), external standards, related plans as relative links.

Lextures-specific defaults

When the spec touches these areas, align with repo conventions (see AGENTS.md):

  • Backend: Go API, Chi router, pgx, JWT auth, server/internal/
  • Frontend: React 19, Vite, TypeScript, Tailwind v4, clients/web/src/
  • Migrations: server/migrations/NNN_descriptive_name.sql
  • Auth: Note JWT scope and role checks on every new route
  • Jobs: Background work via RabbitMQ queue (or in-process fallback locally)
  • Feature flags: Reference OpenFeature-style flags and platformstate when rollout-sensitive
  • Compliance: FERPA for student data; WCAG 2.1 AA for new UI; COPPA for K12 minors

Step 5. Validate

Before delivering, verify every item in references/checklist.md. Fix gaps before presenting.

Delivery format

Present the document in this order:

  1. Summary (3–5 bullets) — what the feature does, severity, effort, key dependencies
  2. Full requirements document — complete markdown, all 19 sections
  3. Suggested next steps — only if useful: save path, related plans to read, open questions needing user input

Examples

User: "Write requirements for instructor bulk email to enrolled students"

Agent: Explores server/internal/ for messaging/email modules and docs/plan/06-communication-collaboration/. Produces a full 19-section plan with FRs for opt-out compliance, rate limits, and audit logging.

User: "Requirements for adding dark mode to the student dashboard"

Agent: Finds existing theme/CSS variables in clients/web/. Status = PARTIAL if theme tokens exist. Non-goals reference system-preference-only scope. ACs cover persistence, contrast ratios, and no flash on load.

Additional resources

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Skeptical analysis and critique of claims, plans, and code without taking anyone at face value. Use when the user says "don't trust me", wants adversarial review, red-teaming, assumption checking, or verification that generated or proposed code is actually correct. Triggers on skepticism, critique everything, verify claims, challenge assumptions, red team, devil's advocate, prove it, smell test.

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

fix-ci

免费

Find failing PR checks, inspect logs or external check links, and apply focused fixes

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

Resolve merge conflicts non-interactively, validate build and tests, and finalize conflict resolution

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

Fetch and summarize review comments from the active pull request

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

how

免费

Use for "how does X work", code walkthroughs before changing something, and placement / ownership / layering questions ("where should this live", "which package owns this", "is this the right layer"). Explains subsystem architecture, runtime flow, onboarding mental models. Can critique architecture. Use why for motivation.

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

implement

免费

Implements a new planned feature

中文简介正在准备中。当前显示原文说明。

StudyDrift/lextures622026年10月14日 更新

查看 StudyDrift 的全部技能

报告此技能的问题