BDK — Broneq Dev Kit
A personal Claude Code plugin packaging reusable dev workflows, skills, agents, and hooks into a single installable unit.
Core design principle: Language-agnostic by default. Skills contain workflow logic only; environment discovery is delegated to STARTUP_INSTRUCTIONS.md and the local project's CLAUDE.md.
Installation
From GitHub directly (recommended until marketplace listing is live):
- Add the BDK marketplace source:
/plugin marketplace add broneq/bdk
- Install:
/plugin install bdk@bdk
Skills
Invoke with /bdk:<skill-name>:
| Skill | Description |
|---|
/bdk:setup | Initialize .bdk/settings.json — run once per project before using other skills |
/bdk:cr | Dynamic code review (3-13 parallel agents based on change size) |
/bdk:commit | Generate conventional commit message from git changes |
/bdk:create-plan | Create TDD-driven implementation plans |
/bdk:execute-plan | Execute a plan with task tracking and verification |
/bdk:subagent-execute-plan | Execute a plan task-by-task with a fresh implementer subagent per task and a single end-of-branch review |
/bdk:verify-plan | Verify a plan against real code before execution |
/bdk:debug | Structured debugging: investigate → failing tests → fix or plan |
/bdk:refactor | Propose object-oriented architecture for complex code |
/bdk:test-driven-development | Rigid TDD cycle: red → green |
/bdk:brainstorming | Design sessions before implementation |
/bdk:create-adr | Generate Architecture Decision Records (MADR format) |
/bdk:save-progress | Checkpoint in-progress work to .bdk/save-progress/ |
/bdk:restore-progress | Resume work from a saved checkpoint |
/bdk:explain-complex-code | Generate architecture docs with Graphviz diagrams |
/bdk:update-docs | Refresh existing architecture docs after code changes |
/bdk:graphviz-docs-compiler | Compile .dot files to SVG and update markdown references |
/bdk:graph-review | Risk-aware code review using knowledge graph — change detection, impact analysis, test coverage |
/bdk:graph-explore | Navigate codebase structure using knowledge graph — stats, architecture, call relationships |
/bdk:graph-debug | Debug issues using graph-powered code navigation — trace call chains, correlate with recent changes |
/bdk:graph-refactor | Safe refactoring using dependency analysis — dead code, rename preview, impact verification |
Agents
Used by skills internally (invoke via subagent_type):
| Agent | Model | Purpose |
|---|
code-reviewer | sonnet | Layer-group deep code review |
implementer | sonnet | End-to-end task implementation (TDD, lint, commit) — used by /bdk:subagent-execute-plan |
fixer | sonnet | Apply specific findings (review, lint, test failures) — used by /bdk:subagent-execute-plan |
explorer | haiku | Fast codebase exploration (CodeGraph → Serena → Grep) |
test-runner | haiku | Run tests, parse and report results |
dead-code-detector | haiku | Find unreachable/unused code |
duplicate-detector | haiku | Find code duplication |
architecture-reviewer | opus | Audit against architectural rules |
static-analyse | haiku | Detect and run project lint/format/type-check |
step-simulator | opus | Dry-run plans with concrete data traces |
log-analyzer | haiku | Parse and summarize error logs |
web-researcher | haiku | Search web for solutions and docs |
Quality Rules
BDK ships language-agnostic rule sets (code-quality, architecture) injected into /bdk:cr and /bdk:create-plan outputs.
Four usage patterns
1. Zero config (recommended for most projects). No settings entry. BDK defaults are used as-is.
2. Extend defaults. Point .bdk/settings.json at a file with project-specific additions:
{
"quality": {
"code-quality": "docs/standards/coding.md"
}
}
affaan-m
ChromeDevTools
f
anthropics