opencode-skills-collection 4.0.19 → 4.0.21

package/bundled-skills/agents-generator/assets/agents-minimal.md

@@ -0,0 +1,60 @@
+# AGENTS.md — Minimal Template
+
+Generate when the user asks for a simple, single-file AGENTS.md following the agents.md standard format (https://agents.md). No YAML frontmatter, no rule files, no ASCII art. Keep it under 30 lines.
+
+Fill `{{PLACEHOLDER}}` markers from project detection. Only include sections whose content is non-empty.
+
+```markdown
+## Setup commands
+
+{{SETUP_COMMANDS}}
+
+## Code style
+
+{{CODE_STYLE}}
+
+## Testing instructions
+
+{{TESTING_INSTRUCTIONS}}
+
+## PR instructions
+
+{{PR_INSTRUCTIONS}}
+```
+
+## Filling rules
+
+### `{{SETUP_COMMANDS}}`
+
+- One bullet per command: install deps, start dev, run build, run lint, format
+- Use the detected package manager (`pnpm install`, `bun install`, etc.)
+- Keep it short — 3-5 bullets max
+
+### `{{CODE_STYLE}}`
+
+- 3-5 project-specific conventions as bullets
+- Only verifiable from config: TypeScript strict, no any, single/double quotes, semicolons, import conventions, UI language
+- Format: `- TypeScript strict mode` not `- We use TypeScript`
+
+### `{{TESTING_INSTRUCTIONS}}`
+
+- What the agent should run before committing
+- The exact test command from package.json
+- One line about adding tests: "- Add or update tests for the code you change"
+- If coverage: mention the command
+
+### `{{PR_INSTRUCTIONS}}`
+
+- Commit format: conventional commits or project-specific
+- What to run before pushing: lint + test + typecheck
+- Branch naming if applicable
+- Attribution rules: "- No AI attribution in commits"
+
+## Omission rules
+
+- Skip any section whose content would be empty
+- Never include architecture descriptions, ASCII diagrams, or stack tables
+- Never include rule file references or `.agents/rules/` mentions
+- If the project has no test runner, skip `## Testing instructions`
+- If the project has no lint setup, omit lint from the PR checklist
+- Keep total file under 30 lines

package/bundled-skills/agents-generator/assets/agents-nested.md

@@ -0,0 +1,73 @@
+# AGENTS.md — Nested Template (monorepo sub-package)
+
+Generate for each sub-package in a monorepo. Lightweight, package-specific. No YAML frontmatter. Format follows the agents.md standard.
+
+Fill `{{PLACEHOLDER}}` markers from the sub-package's package.json and config files.
+
+```markdown
+# {{PACKAGE_NAME}}
+
+{{PACKAGE_DESCRIPTION}}
+
+## Setup commands
+
+- Install: `cd {{PACKAGE_RELATIVE_PATH}} && {{PM}} install`
+- Build: `{{BUILD_COMMAND}}`
+- Dev: `{{DEV_COMMAND}}`
+
+{{TEST_SECTION}}
+
+## Code style
+
+{{CODE_STYLE}}
+```
+
+## Filling rules
+
+### `{{PACKAGE_NAME}}`
+
+From `package.json` `name` field, or directory name if not found.
+
+### `{{PACKAGE_DESCRIPTION}}`
+
+One sentence from `package.json` `description`, or inferred from the package's purpose (shared types, API server, frontend app, etc.).
+
+### `{{PACKAGE_RELATIVE_PATH}}`
+
+Path relative to repo root, e.g. `packages/db` or `apps/dashboard`.
+
+### `{{PM}}`
+
+The root package manager: `pnpm`, `bun`, `npm`, `yarn`.
+
+### `{{BUILD_COMMAND}}`
+
+- Monorepo with turborepo: `npx turbo run build --filter={{PACKAGE_NAME}}`
+- Monorepo with pnpm: `pnpm --filter {{PACKAGE_NAME}} build`
+- If no build script in sub-package: omit or say "`cd {{PACKAGE_RELATIVE_PATH}} && {{PM}} run build`"
+
+### `{{DEV_COMMAND}}`
+
+- Same pattern as build. If sub-package has no dev script, omit.
+
+### `{{TEST_SECTION}}`
+
+```
+## Testing instructions
+- Run tests: `{{TEST_COMMAND}}`
+- Add or update tests for the code you change
+```
+
+Only include if the sub-package has a test script or test config.
+
+### `{{CODE_STYLE}}`
+
+- 2-3 sub-package-specific conventions
+- If sub-package is a shared package: mention exports map, entrypoints, rebuild dependencies
+- If sub-package is an app: mention port, router type, framework
+
+## Omission rules
+
+- Skip the whole nested file if the sub-package has no `package.json`
+- Keep under 20 lines
+- No architecture descriptions, no ASCII diagrams, no tables

package/bundled-skills/agents-generator/assets/architecture.md

@@ -0,0 +1,43 @@
+# Architecture Rules
+
+## {{PROJECT_TYPE_TITLE}}
+
+{{PROJECT_TYPE_DESCRIPTION}}
+
+```text
+{{ARCHITECTURE_DIAGRAM}}
+```
+
+## Stack
+
+| Layer           | Technology | Version |
+| --------------- | ---------- | ------- |
+| {{STACK_TABLE}} |
+
+## {{ROUTING_SECTION_TITLE}}
+
+{{ROUTING_TABLE}}
+
+## Data Flow
+
+```
+{{DATA_FLOW}}
+```
+
+{{SERVER_ACTION_VS_API_SECTION}}
+
+{{RATE_LIMIT_SECTION}}
+
+{{LANGUAGE_SECTION}}
+
+## Generation Rules
+
+- **PROJECT_TYPE_TITLE**: "Single App Structure" if not monorepo, "Monorepo Structure" if monorepo.
+- **ARCHITECTURE_DIAGRAM**: ASCII art of the main directory layout. For monorepos, show apps + shared packages + their relationships. For single apps, show app/ → components/ + lib/ + external service connections.
+- **STACK_TABLE**: Extract from package.json + config files. Each row: layer, technology, exact version.
+- **ROUTING_SECTION_TITLE**: "Routes" for Next.js App Router, "Endpoints" for NestJS/Express, "Routes" generic.
+- **ROUTING_TABLE**: List every route with type (server/client/API) and description. Read from app/ or src/app/.
+- **DATA_FLOW**: ASCII diagram of the main data flow (user action → frontend → backend → database/external → response). Use real function names.
+- **SERVER_ACTION_VS_API_SECTION**: If both server actions and API routes exist, explain which is primary and their differences. If only one, skip the comparison.
+- **RATE_LIMIT_SECTION**: If rate-limit.ts or rate limiting middleware exists, document it with limits and expiry.
+- **LANGUAGE_SECTION**: If the project uses i18n or has UI in a language other than English, document it.

package/bundled-skills/agents-generator/assets/backend.md

@@ -0,0 +1,35 @@
+# Backend Patterns (NestJS)
+
+## Module Architecture (Screaming Architecture)
+
+Each business module is a directory with its own module, controller, and service:
+
+```text
+src/{feature}/
+  {feature}.module.ts       # NestJS module
+  {feature}.controller.ts   # REST routes
+  {feature}-crud.service.ts # Business logic
+  constants.ts              # Module constants
+  __tests__/                # Unit tests
+```
+
+## Rules
+
+- DTOs decorated with validation, validated via global pipe (ZodValidationPipe or class-validator).
+- Dynamic/unmodelable body: receive as `unknown` in controller, cast to narrowest concrete interface in service — never `any`.
+- When splitting a god service, each new service injects only its own dependencies.
+- Services inject the database service from the shared package, not the generated client directly.
+- Role-checking: use enum values directly from the shared package.
+- Use `@nestjs/swagger` decorators for exposed endpoint documentation.
+
+## Error Handling
+
+- Global exception filter catches all unhandled exceptions.
+- Handle database errors: unique constraint, not found, foreign key, relation violation.
+- Typed business errors with code + consistent message.
+- Standard HTTP exceptions for common cases.
+- Auto-log errors >=500 with stack trace.
+
+## Generation Rules
+
+This template is only used when `@nestjs/core` is detected in dependencies. Adapt all paths and module names based on the actual project structure found in `src/`.

package/bundled-skills/agents-generator/assets/claude.md

@@ -0,0 +1,39 @@
+# CLAUDE.md — Template
+
+ONLY generate if Claude Code is detected in the project. Detection signals:
+
+- `.claude/` directory exists
+- `CLAUDE.md` file exists at project root
+- `.claude/skills/` directory exists
+- `claude` appears in devDependencies or as a tool reference
+
+The CLAUDE.md file should be **thin**. Its only job is to point to AGENTS.md. Do not duplicate content.
+
+```markdown
+# CLAUDE.md
+
+See @AGENTS.md
+
+{{CLAUDE_SPECIFIC_SECTION}}
+```
+
+## Filling rules
+
+### `{{CLAUDE_SPECIFIC_SECTION}}`
+
+Only add if the project actually uses Claude-specific features. If none, omit entirely:
+
+- **`.claude/rules/`** — if modular rule files exist: `Additional rules in @.claude/rules/`
+- **`.claude/commands/`** — if custom slash commands exist: `Custom commands: @.claude/commands/`
+- **`.claude/agents/`** — if custom subagents exist: `Custom subagents: @.claude/agents/`
+- **Claude-only behavior**: "Use plan mode for non-trivial tasks", "Use subagents for parallel work"
+
+If none of these apply, CLAUDE.md is just:
+
+```markdown
+# CLAUDE.md
+
+See @AGENTS.md
+```
+
+That's the whole file. Don't add empty sections.

package/bundled-skills/agents-generator/assets/database.md

@@ -0,0 +1,40 @@
+# Database Rules
+
+## ORM: {{ORM_NAME}}
+
+{{ORM_DESCRIPTION}}
+
+## Schema Location
+
+{{SCHEMA_LOCATION}}
+
+{{MIGRATIONS_SECTION}}
+
+## Conventions
+
+{{DB_CONVENTIONS}}
+
+## Relationships and Queries
+
+{{QUERY_PATTERNS}}
+
+## Generation Rules
+
+- **ORM_NAME**: "Prisma", "Drizzle", "Knex", "TypeORM", "Mongoose", or "Raw SQL".
+- **ORM_DESCRIPTION**: Database provider detected (postgresql, sqlite, mysql, mongodb).
+- **SCHEMA_LOCATION**: Where the schema or DB config lives.
+- **MIGRATIONS_SECTION**:
+  - **Prisma**: `prisma migrate dev` for development, `prisma migrate deploy` for production. `prisma generate` to regenerate client.
+  - **Drizzle**: `drizzle-kit generate` for migrations, `drizzle-kit push` for prototyping.
+  - **Knex**: `knex migrate:make`, `knex migrate:latest`.
+- **DB_CONVENTIONS**:
+  - IDs: `uuid` vs `cuid` vs `autoincrement` vs `ObjectId`. Detect from schema.
+  - Timestamps: `createdAt` + `updatedAt` automatically.
+  - Naming: `snake_case` in DB, `camelCase` in code.
+  - Enums: native DB enums vs strings vs TypeScript enums.
+  - Soft deletes: if `deletedAt` field exists.
+- **QUERY_PATTERNS**:
+  - **Prisma**: `prisma.user.findMany({ where, include, select })`. Include vs select rules. Transactions with `prisma.$transaction`.
+  - **Drizzle**: `db.select().from(users).where(eq(...))`. Relations with `relations`.
+  - **Mongoose**: `Model.find()`, `Model.populate()`, `Model.aggregate()`.
+  - **Raw SQL**: Use parameterized queries, never string interpolation.

package/bundled-skills/agents-generator/assets/forms.md

@@ -0,0 +1,56 @@
+# Form Patterns
+
+## Library: {{FORM_LIBRARY}}
+
+{{FORM_LIBRARY_DESCRIPTION}}
+
+## Form Structure
+
+```tsx
+{
+  {
+    FORM_EXAMPLE;
+  }
+}
+```
+
+## Rules
+
+{{FORM_RULES}}
+
+## Validation
+
+{{VALIDATION_SECTION}}
+
+## Errors
+
+{{ERROR_SECTION}}
+
+## Generation Rules
+
+- **FORM_LIBRARY**: "react-hook-form", "formik", "TanStack Form", or "Manual (controlled inputs)".
+- **FORM_LIBRARY_DESCRIPTION**: One sentence about how forms are handled in the project.
+- **FORM_EXAMPLE**: Real example from the project. Read an existing form and adapt. Include: hook import, register/control, handleSubmit, errors, submit handler.
+- **FORM_RULES**:
+  - **react-hook-form**:
+    - `useForm<T>()` with generic type for the form schema.
+    - `register("fieldName")` for simple inputs, `control` + `Controller` for custom components.
+    - `formState: { errors, isSubmitting }` for state.
+    - If using Zod resolver: `zodResolver(schema)` in `useForm`.
+    - Do not mix `useState` with react-hook-form — use `watch()` or `getValues()`.
+  - **formik**:
+    - `useFormik<T>()` or `<Formik>` component.
+    - `validationSchema` for Yup/Zod.
+    - `handleChange`, `handleBlur`, `handleSubmit`.
+  - **Manual**:
+    - `useState` per field.
+    - Manual validation in `handleSubmit`.
+    - No form library.
+- **VALIDATION_SECTION**: If using Zod/Yup/Valibot validation:
+  - Where schemas live (`lib/schemas/`, `src/validations/`).
+  - How they integrate with the form.
+  - Whether they are shared between client and server.
+- **ERROR_SECTION**: How errors are displayed:
+  - Inline below the field.
+  - Toast / sonner for server errors.
+  - Messages in the project's language based on i18n detection.

package/bundled-skills/agents-generator/assets/frontend-patterns.md

@@ -0,0 +1,44 @@
+# Frontend Patterns
+
+## File Structure
+
+```text
+{{FILE_STRUCTURE}}
+```
+
+## Component Rules
+
+{{COMPONENT_RULES}}
+
+## State Management
+
+{{STATE_MANAGEMENT}}
+
+## Toast / Notification System
+
+{{TOAST_SYSTEM}}
+
+## Analytics
+
+{{ANALYTICS_SECTION}}
+
+## Trust Boundaries
+
+{{TRUST_BOUNDARIES}}
+
+## Generation Rules
+
+- **FILE_STRUCTURE**: ASCII tree of relevant source directories (app/, components/, lib/, src/). Include only code directories, not node_modules, .git, etc.
+- **COMPONENT_RULES**:
+  - Next.js App Router: "use client" rule, server/client component split, where each type lives.
+  - Vite/SPA: folder structure, lazy loading.
+  - Tailwind: prohibit inline `style={{}}` (except cases like global-error).
+  - shadcn/ui: atomic components in components/ui/.
+  - UI language for labels/navigation (English/Spanish/etc.) based on project detection.
+- **STATE_MANAGEMENT**:
+  - Zustand/Redux/Jotai: document stores, server hydration patterns.
+  - useState only: document where main state lives.
+  - Shared hooks: list them with their purpose.
+- **TOAST_SYSTEM**: Detect sonner, shadcn toast, custom system, or none. Document the API.
+- **ANALYTICS_SECTION**: If Google Analytics, GTM, Plausible, etc. exists. Document the hook or helper.
+- **TRUST_BOUNDARIES**: Document what the client validates vs what the server validates, what goes through server actions vs client-side.

package/bundled-skills/agents-generator/assets/git-workflow.md

@@ -0,0 +1,61 @@
+# Git Workflow
+
+## Approval Rules
+
+- **NO commits** without the user validating the code first — show the diff.
+- **NO branches** unless explicitly requested.
+- **NO push** unless instructed.
+- Before touching code: show plan (files to touch + proposed commits).
+- After applying: show commands run and results (lint/test/{{QUALITY_TOOL}}).
+
+## Commit Convention
+
+Conventional Commits:
+
+```
+feat: <brief description>
+fix: <brief description>
+refactor: <brief description>
+test: <brief description>
+chore: <brief description>
+docs: <brief description>
+```
+
+Rules:
+
+- One commit per logical change.
+- Short imperative messages.
+- No trailing period.
+- **No "Co-Authored-By" or AI attribution.**
+- {{COMMIT_LANGUAGE}} for commit messages.
+
+{{COMMIT_EXAMPLES}}
+
+## Pre-Commit Verification
+
+Before committing, run:
+
+```
+{{PRE_COMMIT_CYCLE}}
+```
+
+If any step fails, do not commit. Fix and retry.
+
+## Branches
+
+Default base: `{{DEFAULT_BRANCH}}`.
+
+Format:
+
+```
+{{BRANCH_FORMAT}}
+```
+
+## Generation Rules
+
+- **QUALITY_TOOL**: "doctor" if react-doctor is present, otherwise just "lint".
+- **COMMIT_LANGUAGE**: "English" for projects with code in English, "Spanish" if the project uses Spanish in commits.
+- **COMMIT_EXAMPLES**: 2-3 real examples based on the project's typical change types. Read `git log --oneline -5` if available.
+- **PRE_COMMIT_CYCLE**: Shortened version of the full verification cycle.
+- **DEFAULT_BRANCH**: Detect with `git branch --show-current` or `git remote show origin`. Default: "master".
+- **BRANCH_FORMAT**: `feature/<slug>`, `fix/<slug>`, etc. If the project uses issue references (REQ-XXX), include them in the format.

package/bundled-skills/agents-generator/assets/i18n.md

@@ -0,0 +1,50 @@
+# i18n Rules
+
+## Library: {{I18N_LIBRARY}}
+
+{{I18N_DESCRIPTION}}
+
+## Supported Languages
+
+{{LANGUAGES}}
+
+## File Structure
+
+```text
+{{I18N_FILE_STRUCTURE}}
+```
+
+## Usage in Code
+
+```typescript
+{
+  {
+    I18N_USAGE_EXAMPLE;
+  }
+}
+```
+
+## Rules
+
+{{I18N_RULES}}
+
+## Generation Rules
+
+- **I18N_LIBRARY**: "next-intl", "react-i18next", "next-i18next", "Lingui", or "Custom".
+- **I18N_DESCRIPTION**: How i18n works in the project. E.g.: "next-intl with App Router, messages per locale in JSON files, detection via middleware."
+- **LANGUAGES**: List of supported locales. Detect from `i18n.ts`, `next.config.*`, `middleware.ts`, or `messages/{locale}/` folders.
+- **I18N_FILE_STRUCTURE**: Directory tree with translation files.
+- **I18N_USAGE_EXAMPLE**:
+  - **next-intl App Router**: `useTranslations()`, `getTranslations()`, `<NextIntlClientProvider>`.
+  - **next-intl Pages Router**: `getStaticProps` with `useTranslations`.
+  - **react-i18next**: `useTranslation()`, `t("key")`, namespaces.
+  - **next-i18next**: `serverSideTranslations`, `useTranslation`.
+- **I18N_RULES**:
+  - UI uses translations, not hardcoded strings.
+  - Keys in snake_case or camelCase (detect from project).
+  - Interpolation: `t("greeting", { name })` not concatenation.
+  - Date/number formatting with `Intl` or library helpers.
+  - Metadata and SEO also translated (`generateMetadata` with `getTranslations`).
+  - Next.js App Router: middleware handles locale detection + redirect.
+  - Pages Router: `getStaticProps`/`getServerSideProps` load translations.
+  - Translation files nested by feature, not a single flat file.

package/bundled-skills/agents-generator/assets/platform.md

@@ -0,0 +1,54 @@
+# {{PLATFORM_NAME}} Rules — Template
+
+ONLY generate if the corresponding platform is detected in the project. Detection signals per platform below.
+
+This file is wrapped in managed blocks. The agent regenerates content between markers on update; human edits outside markers are preserved.
+
+```text
+<!-- AGENTS-GENERATED-START -->
+# {{PLATFORM_RULES_HEADER}}
+
+{{PLATFORM_SPECIFIC_CONTENT}}
+<!-- AGENTS-GENERATED-END -->
+```
+
+## Platform detection and output
+
+| Platform           | Detection signal                                  | Output file                       | Content mapping                                                          |
+| ------------------ | ------------------------------------------------- | --------------------------------- | ------------------------------------------------------------------------ |
+| **Cursor**         | `.cursorrules` exists or `.cursor/` directory     | `.cursorrules`                    | AGENTS.md sections: Setup commands, Code Style, Hard Rules, Prohibitions |
+| **GitHub Copilot** | `.github/copilot-instructions.md` exists          | `.github/copilot-instructions.md` | Flat instructions: Setup, Code Style, Testing                            |
+| **Gemini CLI**     | `GEMINI.md` exists or `.gemini/` directory        | `GEMINI.md`                       | System prompt style: Overview, Commands, Conventions                     |
+| **Windsurf**       | `.windsurfrules` exists or `.windsurf/` directory | `.windsurfrules`                  | Progressive disclosure: Tech Stack, Rules, Testing                       |
+| **Codex**          | Already covered by AGENTS.md (native support)     | —                                 | AGENTS.md is the primary file                                            |
+
+## Filling rules
+
+### `{{PLATFORM_NAME}}`
+
+The platform name: "Cursor", "GitHub Copilot", "Gemini CLI", "Windsurf".
+
+### `{{PLATFORM_RULES_HEADER}}`
+
+Platform-specific heading:
+
+- Cursor: `# Project Rules: {PROJECT_NAME}`
+- Copilot: `# Copilot Instructions`
+- Gemini: `# Gemini CLI Guidelines`
+- Windsurf: `# Windsurf Rules`
+
+### `{{PLATFORM_SPECIFIC_CONTENT}}`
+
+Adapt from AGENTS.md but use platform-specific format:
+
+- **Cursor**: Progressive disclosure with severity levels. Include BAD/GOOD code examples.
+- **Copilot**: Flat markdown. Keep it short — Copilot has limited context window.
+- **Gemini**: System prompt style. Second person, instructional.
+- **Windsurf**: Progressive disclosure. Include cascade rules.
+
+## Generation rules
+
+1. Only generate for platforms actually detected in the project.
+2. Keep each file under 200 lines — they're companions to AGENTS.md, not replacements.
+3. Include a pointer back to AGENTS.md: `> Full rules in @AGENTS.md`
+4. Use managed blocks (`<!-- AGENTS-GENERATED-START -->` / `<!-- AGENTS-GENERATED-END -->`) to preserve human edits.

package/bundled-skills/agents-generator/assets/sdd-workflow.md

@@ -0,0 +1,46 @@
+# SDD Workflow
+
+Software Design Document — design before implementing.
+
+## Preflight Defaults (cached, do not re-ask)
+
+- **Pace**: {{SDD_PACE}}
+- **Artifacts**: {{SDD_ARTIFACTS}}
+- **Delivery strategy**: {{SDD_DELIVERY}}
+- **Review budget**: {{SDD_REVIEW_BUDGET}} lines
+- **Chain strategy**: {{SDD_CHAIN_STRATEGY}}
+
+## Rules
+
+- Sub-agents can make errors (duplicate blocks, wrong types) — always review their output.
+- After each SDD phase: run the full verification cycle.
+- On archive: verify all tasks marked done, delta specs synced to main specs.
+
+## Post-Apply Verification (MANDATORY)
+
+After `sdd-apply` completes and before advancing to `sdd-verify`, the orchestrator MUST run the verification cycle:
+
+```
+{{VERIFICATION_CYCLE}}
+```
+
+{{MONOREPO_POST_APPLY}}
+{{DOCTOR_POST_APPLY}}
+
+## Spec Verification
+
+Project specs are located at:
+{{SPECS_LOCATIONS}}
+
+Consult these sources before making architectural changes or adding new features.
+
+## Generation Rules
+
+- **SDD_PACE**: "Interactive" by default. "Automatic" if the project uses auto in CI.
+- **SDD_ARTIFACTS**: "both" if OpenSpec + Engram exist, "openspec" if only OpenSpec, "engram" if only Engram.
+- **SDD_DELIVERY**: "ask-on-risk" by default. "exception-ok" if the project prefers large PRs.
+- **SDD_REVIEW_BUDGET**: 400 by default. Adjust based on the project's typical PR size.
+- **SDD_CHAIN_STRATEGY**: "stacked-to-main" if fast merge, "feature-branch-chain" if long feature branches.
+- **MONOREPO_POST_APPLY**: If monorepo, add shared package rebuild steps before the cycle.
+- **DOCTOR_POST_APPLY**: If react-doctor exists, add rule about new issues.
+- **SPECS_LOCATIONS**: List docs/specs/, OpenSpec, relevant Engram topics.

package/bundled-skills/agents-generator/assets/server-actions.md

@@ -0,0 +1,36 @@
+# Server Actions / Backend
+
+## Entry Point
+
+{{ENTRY_POINT_DESCRIPTION}}
+
+## Flow
+
+{{REQUEST_FLOW}}
+
+## {{RESPONSE_TYPE_TITLE}}
+
+{{RESPONSE_TYPE}}
+
+## Error Handling
+
+{{ERROR_HANDLING}}
+
+## Rate Limiting
+
+{{RATE_LIMITING}}
+
+## Generation Rules
+
+- **ENTRY_POINT_DESCRIPTION**:
+  - Server actions: main file(s), exported function(s), parameters.
+  - API Routes: files in app/api/, HTTP methods, parameters.
+  - NestJS: main controllers, guards, DTOs.
+- **REQUEST_FLOW**: Numbered steps of a typical request flow (validation → rate limit → logic → response).
+- **RESPONSE_TYPE_TITLE**: "DownloadResult", "ApiResponse", or the main response type name.
+- **RESPONSE_TYPE**: Interface/type of the response, with comments on optional vs required fields.
+- **ERROR_HANDLING**:
+  - If ErrorCode enum exists: document it with all codes.
+  - If GlobalExceptionFilter exists: document it.
+  - If detectErrorCode or similar exists: document the heuristic.
+- **RATE_LIMITING**: If rate-limit.ts or middleware exists: document mechanism, limits, expiry.