Plan
Solution Architect
Researches the codebase, weighs trade-offs, and turns complex requests into phased implementation plans.
wield:solution-architectWield / Engineering system № 1.5Built for Codex + Claude CodeEst. 2026
A seniorengineeringteam
Plan complex work, build in focused phases, review every edge case, and ship with confidence. One request in. Finished work out.
26 skills · 13 agents · 2 multi-agent workflows
13Specialist agents
Delivery note / 0042Wield crew
Incoming request
Ship team workspaces
01Plansolution-architect
02Buildfullstack-developer
03Reviewcode-reviewer
04Verifyqa-engineer
Final status
Zero blockers
Request inFinished work out
01Deliberately scopedOne clear role per agent
02Context travelsPlans and decisions stay attached
03Verified finishReview and QA close the loop
One toolbelt, end to end
Everything your software team needs to move.
41
Focused skills
What to do
13
Specialized agents
Who does it
3
Dynamic workflows
How teams fan out
1
Shared system
Rules, hooks, and memory
From request to result
AI engineering isn't just about code.It's about outcomes.
Wield turns a loosely defined request into a traceable delivery loop. Each layer has a clear job, and every handoff preserves the context the next specialist needs.
See the complete systemFocused skills
Route the work
Purpose-built playbooks give Claude the right process before it starts changing code.
Specialist agents
Add the right expert
Architects, builders, reviewers, and QA agents each work with a deliberately narrow remit.
Connected workflows
Carry context through
Planning decisions, implementation notes, and verification results stay attached to the job.
Run / 0042Example run
Feature delivery
Team workspace rollout
1. PlanArchitecture mapped
2. Build3 phases complete
3. VerifyReview + QA passed
Recent agent activity
Context syncedsolution-architectMapped four implementation phases09:42
fullstack-developerCompleted the final acceptance criterion10:18
code-reviewerCleared the change with zero blockers10:31
Delivery signal
4
Phases
42
Checks
0
Blockers
Every handoff stays in context.
The final verdict is connected to the plan, the implementation, and the evidence—not reconstructed from scratch.
Shared rules · shared memory · clear ownership
The agent roster
Specialists for every phase of the work.
Wield gives Claude a focused team instead of one oversized prompt. Each agent has a clear job, a scoped toolset, and a place in the engineering loop.
Build
Fullstack Developer
Executes implementation phases across frontend, backend, data, and infrastructure with a clear scope.
wield:fullstack-developerReview
Code Reviewer
Reviews for production readiness and dispatches scouts to find edge cases before they reach a pull request.
wield:code-reviewerDebug
Root-Cause Analyst
Investigates behavior, logs, tests, and data flow to isolate the cause before proposing a fix.
wield:root-cause-analystVerify
QA Engineer
Validates happy paths, failure modes, performance, and build health after meaningful changes.
wield:qa-engineerDesign
UI/UX Designer
Shapes interfaces, responsive systems, interactions, and visual reviews with engineering context intact.
wield:ui-ux-designerOne connected workflow
Plan. Build. Verify. Without losing the thread.
Skills route the work, specialized agents execute it, and shared rules keep every handoff aligned with your codebase.
How it works · 01Auto-advance
Plan with context
The solution architect scouts the codebase, researches the unknowns, and produces a phased plan with explicit boundaries.
Mode
Research
Agent
Architect
Output
Plan
The Wield operating model
More capable because every agent does less.
General assistants are useful for quick edits. Real engineering gets more reliable when planning, implementation, review, and validation each have a specialist.
01
Focused by default
The planning agent plans. The reviewer reviews. Smaller scopes keep the work sharp and the output easier to trust.
02
Parallel when it helps
Research, implementation, and review can fan out across agents with explicit ownership instead of colliding in one context.
03
Verified before it ships
Reviewers, QA, and adversarial workflows challenge the result instead of rubber-stamping the first answer.
Across the whole stack
One context. Every phase. All your tools.
Wield keeps project rules, decisions, and evidence connected from the first scout to the final review.
PLANSHIP
Shared source of truth
Context that survives handoffs
Project rules and plan decisions follow the work instead of getting rediscovered at every phase.
Scoped
Scoped ownership
The right specialist for each phase
Planners plan, builders build, and reviewers challenge the result with a clear boundary around every job.
↗Checks returned42
Verifiable output
Evidence before confidence
Findings come back tied to files, behavior, and checks you can inspect before anything ships.
One connected engineering loop
Every phase leaves the next one a useful artifact, so progress compounds instead of resetting.
- 01Plandecisions captured
- 02Buildcontext preserved
- 03Reviewevidence returned
Your stack, in reach
Works where the work already lives
Code, Git, data, docs, tests, and deployment stay connected to the same project context.
Wield context
The Drop · Pricing
Pick Your Kit
Slim Wield
The open-source, docs-only slim build of Wield. Free forever.
Free
- Docs-focused plugin — core app docs only
- Community updates as the project grows
Engineer Kit
Ship agent-grade engineering workflows from day one.
$39.99$399.9990% Off
- 12 production guides across the engineer surface
- CLI + plugin catalog access
- Lifetime updates to the engineer track
Teams & Enterprise
Team and enterprise access is coming soon and isn't supported yet.
Coming Soon
- Shared access for engineering teams
- Centralized rollout and workspace controls
- Organization billing and support
The signal / 305
Used by
305
engineers
A growing crew of engineers uses Wield to turn a task into a focused, reviewed, and verified delivery loop.
Clear by design
Questions, answered.
What Wield installs, how the team works, and where the optional workflows fit.
What exactly is Wield?
Wield supports Codex and Claude Code. It adds 26 slash-command skills, 13 specialized agents, shared engineering rules, and a small set of unobtrusive hooks to your existing setup.
How do I install it?
Wield supports Codex and Claude Code. For Claude Code, add the marketplace and install the plugin with two lines: /plugin marketplace add wield-ai/engineering-plugin, then /plugin install wield@wield.
How do skills and agents relate?
Skills describe what to do; agents are who does it. A skill such as /wield:implementation-planning routes the work and can dispatch the solution-architect or research agents behind the scenes.
Do I need to call each agent myself?
Usually not. You invoke a skill such as /wield:feature-building or /wield:code-audit, and Wield dispatches the specialist that fits the job. Agents are also available directly when you need finer control.
What are the three multi-agent workflows?
/ultra-review runs parallel review lenses, /security-sweep red-teams the codebase, and /bug-sweep hunts latent defects. Each workflow merges, deduplicates, and adversarially verifies its findings.
Are the workflows installed with the plugin?
Not automatically, because Claude Code plugins cannot bundle workflows. Wield includes an install script that copies them into .claude/workflows; they require Claude Code v2.1.154+ with workflows enabled.
Does Wield require configuration?
No. Every skill has sensible defaults and works immediately after installation. Optional settings live in .wield.json when you want to tune models, plan naming, or hook behavior.
Is there a subscription?
The Engineer Kit is a one-time purchase with lifetime access to its updates. Slim Wield is the free, open-source docs-focused build.
What does Wield require?
Codex or Claude Code and Git are required. Node.js is recommended for the privacy and context guard hooks; without it, those hooks fail open while the rest of Wield continues to work.
Looking for the technical details? Read the complete docs
Your team is ready