| name | vibe-kanban-brainstorming |
| description | Plans and scaffolds Anthropic API endpoints as Terraform resources/data sources for Vibe Kanban. TRIGGER when: user says "implement GitHub issues", references a range of GitHub issues to implement (e.g. "issues 52 to 57"), asks to parallelize implementation of multiple resources/data sources, or says "implement all [API name] APIs". Creates Vibe Kanban issues, workspaces, and fires parallel agents — one per resource/data source.
|
| model | opus |
vibe-kanban-brainstorming
Concepts
- Issues — kanban tasks; this skill creates them.
- Workspaces — isolated git worktrees, one per issue. The human operator creates only the orchestration workspace manually; the orchestration agent creates all implementation workspaces automatically.
- Orchestration issue — one required per batch; monitors parallel work on
main, no code output. Implementation issues are blocking it so it stays open until all work is merged.
Workflow
Step 1 — Discover the API
Fetch https://platform.claude.com/docs/en/api/overview and all relevant endpoint pages in parallel. Collect HTTP method, path, request/response fields.
Endpoint → Terraform mapping:
POST + GET /{id} + DELETE /{id} → resource (ForceNew)
POST + GET /{id} + PATCH /{id} + DELETE /{id} → resource (with Update)
GET /{id} only → data source (single)
GET / → data source (list)
Step 2 — Create Vibe Kanban issues
One mcp__vibe_kanban__create_issue per resource/data source plus one orchestration issue, all in a single parallel message.
All issues must be top-level: never set parent_issue_id.
Priorities: orchestration high, resources high, data sources medium.
Orchestration issue body:
Track parallel implementation of <API name> resources and data sources.
## Workspaces to monitor
- [ ] <Issue title A>
- [ ] <Issue title B>
## Agent instructions
Do NOT write code. Your job:
1. Create a workspace for each implementation issue above (the human has only created yours).
2. Monitor sibling workspaces; report progress and flag inconsistencies.
3. Close this issue once all sibling PRs are merged.
Implementation issue body: use the template at the bottom of this file.
Step 3 — Create issue relationships
Call mcp__vibe_kanban__create_issue_relationship for all pairs in a single parallel message:
- Every implementation issue → orchestration issue:
blocking
- Resource ↔ its data source(s):
related
- Single data source ↔ list data source:
related
Step 4 — Create GitHub issues
First check for existing issues to avoid duplicates:
gh issue list --search "<resource name>" --json number,title,url
For each implementation issue only (orchestration has no GitHub issue), create via gh CLI in a single parallel message:
gh issue create --title "<title>" --body "<body>"
Do not create feat/ branches here — the agent's /ship skill creates the branch automatically when it opens a PR.
Step 5 — Cross-link
Update all Vibe Kanban issues in a single parallel message:
- Implementation issues: prepend
GitHub Issue: <url>
- Orchestration issue: prepend a list of all GitHub issue URLs it tracks.
Step 6 — Create implementation workspaces
Call mcp__vibe_kanban__start_workspace for every implementation issue in a single parallel message. Always pass "branch": "main" — vibe-kanban auto-generates a vk/ branch internally.
mcp__vibe_kanban__start_workspace({
name: "<resource-or-datasource-name>",
executor: "CLAUDE_CODE",
issue_id: "<vk-issue-id>",
repositories: [{"repo_id": "<repo-id>", "branch": "main"}]
})
After all calls return a workspace_id, call mcp__vibe_kanban__list_sessions for each workspace to retrieve the session ID.
If any start_workspace call returns an error: do not call run_session_prompt for that workspace — delete it and recreate it before proceeding.
Step 7 — Fire implementation prompts
Call mcp__vibe_kanban__run_session_prompt for all sessions in a single parallel message. Each prompt must include the /terraform-provider skill invocation followed by the full spec (API endpoints, SDK methods, Terraform schema, files to create).
Step 8 — Inform the user
Report a summary table of issues, GitHub issue URLs, and workspace IDs. Remind: "Each agent will open a draft PR via /ship when it's ready."
Invariants
- All issues are top-level. Never set
parent_issue_id on any issue.
- One orchestration issue per batch. No branch, no PR, no commits. Never skip it.
start_workspace always uses branch: "main". Never pass a feat/ or custom branch name — vibe-kanban generates its own vk/ branch. Passing a non-existent branch causes a 400 error and leaves the worktree uninitialised, making subsequent run_session_prompt calls fail.
- One GitHub issue per implementation ticket. URL written back into the Vibe Kanban description.
- One PR per ticket. Never bundle multiple tickets. The agent's
/ship skill creates the branch and PR.
- No Vibe Kanban IDs outside Vibe Kanban (not in commits, PR titles, or PR bodies — use
#N).
- Branch isolation. Each agent commits only to its own worktree branch.
Implementation issue template
GitHub Issue: <url>
<One-line summary.>
## API
- `METHOD /v1/<path>` — description
## SDK methods
- `client.<Service>.<Method>(ctx, params)` → `ResponseType`
## Terraform schema
| Attribute | Type | R/W |
|---|---|---|
| `id` | string | Computed |
## Implementation
Follow the `terraform-provider-implementer` agent conventions (files, tests, make, review).
Example
User: "I'd like to implement /v1/environments APIs"
- Fetch API docs in parallel → identify:
anthropic_environment resource, anthropic_environment data source, anthropic_environments data source.
- Create 4 Kanban issues in parallel: 1 orchestration + 3 implementation.
- Create 5 relationships in parallel: 3× impl
blocking orch, resource related single DS, single DS related list DS.
- Create 3 GitHub issues in parallel (implementation only; no branches).
- Update 4 Kanban issues with GitHub links in parallel.
- Start 3 workspaces in parallel; retrieve session IDs.
- Fire 3 implementation prompts in parallel.
- Report summary.