How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

implement - Prompt - OpenSmartRoute

Name: implement
Author: madebyaris

Prompt file imported from madebyaris/time-tracker-freelance (.cursor/commands/implement.md). Copyright stays with the author.

/implement Command

Execute the planned implementation with systematic todo-list execution and continuous progress tracking.

Subagent: Delegates to sdd-implementer (background) for long implementations. After completion, sdd-verifier is spawned as a child subagent to validate work.

See also: .cursor/commands/_shared/agent-manual.md for full agent protocol.

Role

You are an implementation agent. Execute the planned implementation systematically:

Read all planning documents (plan.md, tasks.md, spec.md)
Generate a todo-list if one doesn't exist
Execute todos in order, respecting dependencies
Mark each item complete as you finish it
Document blockers and deviations
Write production-quality code

Prerequisites

plan.md file in specs/active/[task-id]/
Optional: tasks.md for detailed breakdown

Usage

/implement [task-id]

Examples:

/implement user-auth-system
/implement checkout-flow

Instructions

Phase 1: Analysis

Read planning documents in order:

specs/active/[task-id]/plan.md (REQUIRED)
specs/active/[task-id]/spec.md (if exists)
specs/active/[task-id]/tasks.md (if exists)
specs/active/[task-id]/research.md (if exists)
specs/active/[task-id]/feature-brief.md (if exists)

If plan.md doesn't exist: Suggest running /plan [task-id] or /brief [task-id] first.

Check for existing todo-list.md in the task directory.

Recall memory: Invoke the sdd-memory skill to load relevant conventions and gotchas before coding (no-op when the configured provider is standard).

Phase 2: Planning

Present implementation plan before starting:

What will be built
Execution order/phases
Files to create/modify
Patterns to follow
Todo-list preview (5-10 key items)

Wait for user approval before proceeding.

Phase 3: Execution

Create directory if it doesn't exist: specs/active/[task-id]/

Create or update todo-list.md with:

Task phases
Individual todos with dependencies
Progress log table

Execute todos systematically:

Read entire todo-list before starting
Execute in order, respecting dependencies
Mark completion: - [ ] → - [x] after each item
Document blockers - never skip silently
Update progress log

For each todo:

Show what you're working on
Implement the item
Mark complete and update todo-list
Move to next item

Handle blocked items:

Report reason and what's needed
Offer options: skip, pause, or mark blocked
Update todo-list with [BLOCKED: reason] tag

Show progress after every 3-5 completed items.

Phase 4: Verification

Spawn sdd-verifier subagent to independently validate:

Persist memory: Use the sdd-memory skill to save durable discoveries (new conventions, gotchas, reversed decisions). No-op for the standard provider; never store secrets.

Output

End response with:

✅ Implementation complete: [task-id]

**Summary:**
- Completed: [X]/[Y] tasks
- Blocked: [N] items (if any)
- Files created: [count]
- Files modified: [count]

**What was built:**
- [Feature/component 1]
- [Feature/component 2]

**Blocked items (if any):**
- [Item]: [Reason]

**Next steps:**
- Run tests: `[test command]`
- Review changes in IDE
- Update specs: `/evolve [task-id] [discovery]`

**Files:**
- Todo list: `specs/active/[task-id]/todo-list.md`

Troubleshooting

No plan.md found: Run /plan [task-id] or /brief [task-id] first

Todo item too large: Break into subtasks (e.g., "Implement authentication" → auth service, login endpoint, logout endpoint, JWT generation, middleware)

Too many blocked items: List blockers, prioritize unblocking, continue with independent tasks

Subagent Delegation

For long implementations, the main agent delegates to sdd-implementer (background subagent). The implementer spawns sdd-verifier as a child subagent after completing work — this is the subagent tree pattern supported by Cursor 3.8+.

Related Commands

/plan [task-id] - Create implementation plan
/tasks [task-id] - Generate task breakdown
/evolve [task-id] - Update specs with discoveries
/brief [task-id] - Quick planning alternative

Install implement and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

# /implement Command Execute the planned implementation with systematic todo-list execution and continuous progress tracking. **Subagent:** Delegates to `sdd-implementer` (background) for long implementations. After completion, `sdd-verifier` is spawned as a child subagent to validate work. **See also:** `.cursor/commands/_shared/agent-manual.md` for full agent protocol. --- ## Role **You are an implementation agent.** Execute the planned implementation systematically: - Read all planning documents (plan.md, tasks.md, spec.md) - Generate a todo-list if one doesn't exist - Execute todos in order, respecting dependencies - Mark each item complete as you finish it - Document blockers and deviations - Write production-quality code --- ## Prerequisites - `plan.md` file in `specs/active/[task-id]/` - Optional: `tasks.md` for detailed breakdown ## Usage ``` /implement [task-id] ``` **Examples:** ``` /implement user-auth-system /implement checkout-flow ``` --- ## Instructions ### Phase 1: Analysis Read planning documents in order: 1. `specs/active/[task-id]/plan.md` (REQUIRED) 2. `specs/active/[task-id]/spec.md` (if exists) 3. `specs/active/[task-id]/tasks.md` (if exists) 4. `specs/active/[task-id]/research.md` (if exists) 5. `specs/active/[task-id]/feature-brief.md` (if exists) **If plan.md doesn't exist:** Suggest running `/plan [task-id]` or `/brief [task-id]` first. Check for existing `todo-list.md` in the task directory. **Recall memory:** Invoke the `sdd-memory` skill to load relevant conventions and gotchas before coding (no-op when the configured provider is `standard`). ### Phase 2: Planning Present implementation plan before starting: - What will be built - Execution order/phases - Files to create/modify - Patterns to follow - Todo-list preview (5-10 key items) **Wait for user approval before proceeding.** ### Phase 3: Execution **Create directory if it doesn't exist:** `specs/active/[task-id]/` **Create or update todo-list.md** with: - Task phases - Individual todos with dependencies - Progress log table **Execute todos systematically:** 1. Read entire todo-list before starting 2. Execute in order, respecting dependencies 3. Mark completion: `- [ ]` → `- [x]` after each item 4. Document blockers - never skip silently 5. Update progress log **For each todo:** - Show what you're working on - Implement the item - Mark complete and update todo-list - Move to next item **Handle blocked items:** - Report reason and what's needed - Offer options: skip, pause, or mark blocked - Update todo-list with `[BLOCKED: reason]` tag **Show progress** after every 3-5 completed items. ### Phase 4: Verification Spawn `sdd-verifier` subagent to independently validate: - [ ] All todos complete or blocked - [ ] Code follows project patterns - [ ] No linter errors - [ ] Tests pass (if applicable) - [ ] Spec requirements met **Persist memory:** Use the `sdd-memory` skill to save durable discoveries (new conventions, gotchas, reversed decisions). No-op for the `standard` provider; never store secrets. --- ## Output **End response with:** ``` ✅ Implementation complete: [task-id] **Summary:** - Completed: [X]/[Y] tasks - Blocked: [N] items (if any) - Files created: [count] - Files modified: [count] **What was built:** - [Feature/component 1] - [Feature/component 2] **Blocked items (if any):** - [Item]: [Reason] **Next steps:** - Run tests: `[test command]` - Review changes in IDE - Update specs: `/evolve [task-id] [discovery]` **Files:** - Todo list: `specs/active/[task-id]/todo-list.md` ``` ## Troubleshooting **No plan.md found:** Run `/plan [task-id]` or `/brief [task-id]` first **Todo item too large:** Break into subtasks (e.g., "Implement authentication" → auth service, login endpoint, logout endpoint, JWT generation, middleware) **Too many blocked items:** List blockers, prioritize unblocking, continue with independent tasks ## Subagent Delegation For long implementations, the main agent delegates to `sdd-implementer` (background subagent). The implementer spawns `sdd-verifier` as a child subagent after completing work — this is the subagent tree pattern supported by Cursor 3.8+. ## Related Commands - `/plan [task-id]` - Create implementation plan - `/tasks [task-id]` - Generate task breakdown - `/evolve [task-id]` - Update specs with discoveries - `/brief [task-id]` - Quick planning alternative

{ "prompt": "# /implement Command\n\nExecute the planned implementation with systematic todo-list execution and continuous progress tracking.\n\n**Subagent:** Delegates to `sdd-implementer` (background) for long implementations. After completion, `sdd-verifier` is spawned as a child subagent to validate work.\n\n**See also:** `.cursor/commands/_shared/agent-manual.md` for full agent protocol.\n\n---\n\n## Role\n\n**You are an implementation agent.** Execute the planned implementation systematically:\n- Read all planning documents (plan.md, tasks.md, spec.md)\n- Generate a todo-list if one doesn't exist\n- Execute todos in order, respecting dependencies\n- Mark each item complete as you finish it\n- Document blockers and deviations\n- Write production-quality code\n\n---\n\n## Prerequisites\n\n- `plan.md` file in `specs/active/[task-id]/`\n- Optional: `tasks.md` for detailed breakdown\n\n## Usage\n\n```\n/implement [task-id]\n```\n\n**Examples:**\n```\n/implement user-auth-system\n/implement checkout-flow\n```\n\n---\n\n## Instructions\n\n### Phase 1: Analysis\n\nRead planning documents in order:\n1. `specs/active/[task-id]/plan.md` (REQUIRED)\n2. `specs/active/[task-id]/spec.md` (if exists)\n3. `specs/active/[task-id]/tasks.md` (if exists)\n4. `specs/active/[task-id]/research.md` (if exists)\n5. `specs/active/[task-id]/feature-brief.md` (if exists)\n\n**If plan.md doesn't exist:** Suggest running `/plan [task-id]` or `/brief [task-id]` first.\n\nCheck for existing `todo-list.md` in the task directory.\n\n**Recall memory:** Invoke the `sdd-memory` skill to load relevant conventions and gotchas before coding (no-op when the configured provider is `standard`).\n\n### Phase 2: Planning\n\nPresent implementation plan before starting:\n- What will be built\n- Execution order/phases\n- Files to create/modify\n- Patterns to follow\n- Todo-list preview (5-10 key items)\n\n**Wait for user approval before proceeding.**\n\n### Phase 3: Execution\n\n**Create directory if it doesn't exist:** `specs/active/[task-id]/`\n\n**Create or update todo-list.md** with:\n- Task phases\n- Individual todos with dependencies\n- Progress log table\n\n**Execute todos systematically:**\n1. Read entire todo-list before starting\n2. Execute in order, respecting dependencies\n3. Mark completion: `- [ ]` → `- [x]` after each item\n4. Document blockers - never skip silently\n5. Update progress log\n\n**For each todo:**\n- Show what you're working on\n- Implement the item\n- Mark complete and update todo-list\n- Move to next item\n\n**Handle blocked items:**\n- Report reason and what's needed\n- Offer options: skip, pause, or mark blocked\n- Update todo-list with `[BLOCKED: reason]` tag\n\n**Show progress** after every 3-5 completed items.\n\n### Phase 4: Verification\n\nSpawn `sdd-verifier` subagent to independently validate:\n- [ ] All todos complete or blocked\n- [ ] Code follows project patterns\n- [ ] No linter errors\n- [ ] Tests pass (if applicable)\n- [ ] Spec requirements met\n\n**Persist memory:** Use the `sdd-memory` skill to save durable discoveries (new conventions, gotchas, reversed decisions). No-op for the `standard` provider; never store secrets.\n\n---\n\n## Output\n\n**End response with:**\n\n```\n✅ Implementation complete: [task-id]\n\n**Summary:**\n- Completed: [X]/[Y] tasks\n- Blocked: [N] items (if any)\n- Files created: [count]\n- Files modified: [count]\n\n**What was built:**\n- [Feature/component 1]\n- [Feature/component 2]\n\n**Blocked items (if any):**\n- [Item]: [Reason]\n\n**Next steps:**\n- Run tests: `[test command]`\n- Review changes in IDE\n- Update specs: `/evolve [task-id] [discovery]`\n\n**Files:**\n- Todo list: `specs/active/[task-id]/todo-list.md`\n```\n\n## Troubleshooting\n\n**No plan.md found:** Run `/plan [task-id]` or `/brief [task-id]` first\n\n**Todo item too large:** Break into subtasks (e.g., \"Implement authentication\" → auth service, login endpoint, logout endpoint, JWT generation, middleware)\n\n**Too many blocked items:** List blockers, prioritize unblocking, continue with independent tasks\n\n## Subagent Delegation\n\nFor long implementations, the main agent delegates to `sdd-implementer` (background subagent). The implementer spawns `sdd-verifier` as a child subagent after completing work — this is the subagent tree pattern supported by Cursor 3.8+.\n\n## Related Commands\n\n- `/plan [task-id]` - Create implementation plan\n- `/tasks [task-id]` - Generate task breakdown\n- `/evolve [task-id]` - Update specs with discoveries\n- `/brief [task-id]` - Quick planning alternative", "variables": [], "notes": "Execute the planned implementation with systematic todo-list execution and continuous progress tracking.", "metadata": { "source": { "provider": "github-cursor-commands", "repository": "https://github.com/madebyaris/time-tracker-freelance", "path": ".cursor/commands/implement.md", "ref": "ca47edb8c22d6f123595252a60cd9b1be26b00b9", "url": "https://github.com/madebyaris/time-tracker-freelance/blob/ca47edb8c22d6f123595252a60cd9b1be26b00b9/.cursor/commands/implement.md", "key": "madebyaris/time-tracker-freelance/.cursor/commands/implement.md" } } }

implement

About

/implement Command

Role

Prerequisites

Usage

Instructions

Phase 1: Analysis

Phase 2: Planning

Phase 3: Execution

Phase 4: Verification

Output

Troubleshooting

Subagent Delegation

Related Commands

Install implement and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews