From template-bridge
Enforces structured dev workflow using Beads for task tracking/dependencies, Superpowers for brainstorm/plan/TDD/review, git worktrees for isolation. Triggers at session start, before tasks, or unclear workflow.
npx claudepluginhub maslennikov-ig/template-bridge --plugin template-bridgeThis skill uses the workspace's default tool permissions.
| System | Role | Type |
Guides using Beads (bd), a distributed git-backed graph issue tracker for managing tasks, tracking dependencies, AI agent workflows, and multi-branch git development.
Guides Beads (bd) usage for git-backed task management, dependency tracking, AI agent workflows, and multi-branch parallel development.
Guides task implementation using Conductor's TDD workflow, phase checkpoints, git commits, and verification protocols.
Share bugs, ideas, or general feedback.
| System | Role | Type |
|---|---|---|
| Superpowers | Development methodology (HOW to work) | Plugin, rigid skills |
| Beads | Task tracking (WHAT to work on) | Plugin, bd CLI |
| Template Bridge | Specialist agents (WHO helps) | On-demand from claude-code-templates |
Task arrives
│
├─ 1. EPIC: bd create -t epic "High-level goal"
│ (container — tracks intent, holds lock, preserves context)
│
├─ 2. BRAINSTORM: superpowers:brainstorming (design before code)
│
├─ 3. PLAN: superpowers:writing-plans (decompose into 2-5 min tasks)
│
├─ 4. SUB-TASKS: Create concrete tasks from the plan:
│ bd create "Sub-task 1" -t task
│ bd dep add sub-task-1 epic-id --type parent-child
│ bd dep add sub-task-2 sub-task-1 (if sequential — blocks)
│
├─ 5. ISOLATE: superpowers:using-git-worktrees (non-trivial work)
│
├─ 6. IMPLEMENT: bd ready → pick from Ready Front → bd update <id> --claim
│ For each sub-task:
│ ├─ TDD: RED → verify fail → GREEN → verify pass → REFACTOR
│ ├─ Commit after each green
│ ├─ superpowers:requesting-code-review
│ └─ bd close <sub-id> --reason "Done"
│ (closing unblocks next Ready Front automatically)
│
├─ 7. VERIFY: superpowers:verification-before-completion
│
├─ 8. FINISH: superpowers:finishing-a-development-branch
│
└─ 9. CLOSE: bd close <epic-id> --reason "Done"
Four dependency types (only blocks affects bd ready):
| Type | Effect on bd ready | Use when |
|---|---|---|
| blocks | YES — blocked task hidden from bd ready | Sequential work, technical prerequisites |
| parent-child | No | Epic → sub-task structure |
| related | No | Connected but independent work |
| discovered-from | No | Side-quest found during implementation |
Direction rule: Think "X needs Y" → bd dep add X Y
Ready Fronts: As sub-tasks close, blocked work automatically becomes ready.
Use bd ready to always see what can be worked on NOW.
Side quests during implementation:
bd create "Found: missing input validation" -t bug
bd dep add new-id current-id --type discovered-from
No production code without a failing test first. Ever.
Invoke superpowers:systematic-debugging BEFORE any fix attempt:
Before claiming ANYTHING done, invoke superpowers:verification-before-completion:
"Evidence before claims, always."
bd ready → find available tasks from current Ready Frontbd update <id> --claim → atomic start (marks in_progress + lock)bd close <id> --reason "..." → mark done, auto-unblocks dependentsbd create "Issue" -t bug + bd dep add new-id current-id --type discovered-frombd list --status in_progress → bd show <id> → read notesCheck template-bridge:template-catalog when task needs a specialist NOT in:
Install fresh: npx claude-code-templates@latest --agent {category}/{name} --yes
Before implementing with ANY external library, framework, or SDK:
mcp__context7__resolve-library-id → mcp__context7__query-docs
When: Next.js, Supabase, React, Tailwind, Zod, BullMQ, Qdrant, LangChain — any library. Who: Main agent, subagents, and skills — everyone queries Context7. Why: Claude's training data is stale. APIs change. Context7 gives current docs.
Never write implementation code based on memory alone. Always verify with Context7 first.
| Skill | When | Type |
|---|---|---|
brainstorming | Before ANY creative/feature work | Rigid |
writing-plans | After brainstorm approval | Rigid |
test-driven-development | During ALL implementation | Rigid |
systematic-debugging | Any bug/test failure | Rigid |
verification-before-completion | Before claiming done | Rigid |
using-git-worktrees | Non-trivial feature work | Flexible |
requesting-code-review | After each task | Flexible |
receiving-code-review | When getting feedback | Rigid |
subagent-driven-development | Executing plans, independent tasks | Flexible |
executing-plans | Following pre-written plans | Flexible |
finishing-a-development-branch | Tests pass, ready to merge | Rigid |
dispatching-parallel-agents | 2+ independent problems | Flexible |