By lookatitude
A team-of-specialists plugin for Claude Code: 14 domain specialists across engineering, content, and commercial groups, a brainstorm-plan-execute-review-verify-reflect spine, a categorized wiki with decision capture, and a self-evolution loop with shadow-mode gating.
Security audit of installed Guild scripts — SHA-256 hash every file, flag network/filesystem access patterns, and record results. Per guild-plan.md §13.1 and §15.1 #12.
Kick off skill evolution — runs the eval loop, shadow mode, and flip-gating promotion pipeline. Manual mode pauses at the promotion gate for user review; --auto runs unattended if the gate criteria are met without regressions. Per guild-plan.md §13.1.
Revert a skill to a previous version by snapshotting the current state as vN+1 and restoring vN (or vN-n). Non-destructive — every rollback is itself a versioned artifact. Per guild-plan.md §13.1 and §11.3.
Show usage stats — task count, completion rate, flip counts, top-used skills, top-requested specialists, open reflection backlog, and audit drift summary. Pure read; never writes. Per guild-plan.md §13.1.
Team operations — propose a new team, show the current team, or edit it. Per guild-plan.md §13.1. Dispatches to guild-team-compose skill. `edit --allow-larger` lifts the 6-specialist cap from §7.2.
Owns systems design, architectural tradeoff analysis, and ADR authoring. Produces design docs, tradeoff matrices, and Architecture Decision Records — not implementation code. TRIGGER for "design this system", "system design", "architecture review", "architect this", "pick between X and Y", "tradeoff analysis", "evaluate options", "write an ADR", "architectural decision". Implied specialist on any multi-component build per guild-plan.md §7.2. DO NOT TRIGGER for: implementation coding (backend owns APIs, data layer, service integration); pre-decision research, paper digests, vendor comparison (researcher supplies inputs, architect decides); test authoring, coverage, property/snapshot/flaky work (qa); deploy, CI/CD, IaC, observability (devops); security audits, threat models, CVE scans (security); iOS/Android/RN implementation (mobile); skill authoring, hook engineering — dev-team; web frontend implementation (frontend); UI/visual design (no dedicated visual-design specialist yet — flag as followup).
Owns API contracts, data layer, migrations, service integrations, queue consumers, and worker jobs. TRIGGER for "design the API", "add endpoint", "schema change", "write migration", "integrate with service X", "queue consumer", "worker job", REST/GraphQL contract work, database table/column changes, data-access code, external API clients, job scheduling. DO NOT TRIGGER for: systems architecture and cross-component design (architect — backend implements after architect's contract sketch); pre-decision research, paper digests, vendor comparison (researcher); test authoring, coverage, property/snapshot/flaky (qa — backend writes pinning tests, qa owns suite shape); deploy, CI/CD, IaC, observability, release pipelines (devops); security audits, threat models, auth-flow review, CVE scans, secrets scans (security — backend writes auth code, security reviews); iOS/Android/RN client implementation (mobile); web frontend (frontend); content, marketing, copy, SEO (writing/commercial groups).
Owns long-form blog writing, product microcopy, lifecycle email sequences, and voice-guide authoring. TRIGGER for "write a blog post", "draft an article", "product microcopy", "button label", "empty-state copy", "onboarding copy", "welcome series", "transactional email copy", "define brand voice", "rewrite this landing page copy". DO NOT TRIGGER for: technical docs, API docs, user manuals, tutorials, changelogs, release notes (technical-writer); social posts, threads, captions, content calendars (social-media); SEO keyword research, on-page optimization, technical audits, internal linking (seo); positioning, messaging frameworks, GTM, launch plans, campaign briefs, A/B variant programs (marketing); cold outreach, prospecting, discovery scripts, proposals, follow-up sequences (sales); code, tests, infra, deployment, security work.
Owns CI/CD pipelines, infrastructure-as-code, observability stacks, release/rollout, and incident runbooks. Produces pipeline configs, Terraform/Pulumi/CDK, dashboards, alerts, SLOs, runbooks — not application code. TRIGGER for "deploy", "pipeline", "CI", "CD", "release", "rollout", "canary", "blue/green", "infra", "Terraform", "Kubernetes manifest", "observability", "dashboard", "alert", "SLO", "SLI", "error budget", "runbook", "on-call", "incident". DO NOT TRIGGER for: systems architecture (architect); application code, API, migrations, integrations (backend); test strategy or authoring (qa); threat models, CVE scans, auth-flow review (security — devops wires scanners, security defines rules); mobile build configs (mobile owns EAS/Fastlane); research briefs or vendor benchmarks (researcher — devops picks the pipeline, researcher surveys options); skill authoring, hook engineering under .claude/agents/.
Owns web frontend implementation across React, Vue, Svelte, and Solid — component authoring, state management, client-side routing, bundler configuration (Vite/Webpack/esbuild), styling systems (Tailwind/CSS modules/vanilla), accessibility, and frontend performance. Produces working frontend code, not visual design. TRIGGER for "build the React app", "Vite + React", "frontend component", "build a page in <framework>", "state management", "client-side routing", "Vite config", "Tailwind setup", "responsive layout", "a11y", "accessibility", "Lighthouse", "Core Web Vitals fix", "lazy load", "code splitting", "frontend bundle size", "React Query", "Redux", "Zustand", "Pinia", "Svelte store", "design-system implementation", "charts in React". DO NOT TRIGGER for: cross-system architecture (architect — frontend implements after architect's contract sketch); API contracts, data layer, server endpoints (backend — frontend consumes the contract); test strategy, suite shape, property/snapshot/flaky work (qa — frontend writes its own component tests, qa owns suite shape); iOS/Android/React Native client implementation (mobile — distinct runtime); CI/CD, deployment, IaC, observability (devops); UI microcopy and product strings (copywriter — frontend wires the strings in); technical SEO audits (seo diagnoses, frontend implements the fix); positioning, GTM, campaign work (marketing); long-form content, voice guides (copywriter); skill authoring, hook engineering — dev-team.
Authors Terraform / Pulumi / CDK for a cloud component — compute, networking, storage, IAM — with state backend, drift detection, and module boundaries. Output: IaC source files plus backend/state configuration and a module-tree note. Pulled by the `devops` specialist. TRIGGER: "write Terraform for X", "provision the infra for X", "author the CDK stack for X", "Pulumi program to set up X", "create the IaC module for X", "stand up the cloud resources for X". DO NOT TRIGGER for: CI/CD workflow / pipeline YAML (use `devops-ci-cd-pipeline`), monitoring / dashboards / alerts (use `devops-observability-setup`), runbook authoring (use `devops-incident-runbook`), high-level architecture across services (architect-systems-design), app-level config inside a service (backend group), secret rotation policy (security-secrets-scan).
Wires metrics, logs, traces, SLO dashboards, and alerts for a service. Output: dashboard-as-code (Grafana / Datadog / CloudWatch JSON), alert rules, and the SLO → alert mapping. Pulled by the `devops` specialist. TRIGGER: "add observability for X", "wire metrics and alerts for X", "set up the dashboard for X", "define SLOs and alerts for X", "instrument X with tracing", "what should we alert on for X". DO NOT TRIGGER for: CI/CD pipeline setup (use `devops-ci-cd-pipeline`), provisioning the underlying infra (use `devops-infrastructure-as-code`), writing the runbook the alert links to (use `devops-incident-runbook`), app-level logging code (backend group), security event monitoring as a compliance deliverable (security-threat-modeling), test flakiness telemetry (qa-flaky-test-hunter).
Authors A/B (or multivariate) copy variants with a testable hypothesis, a single changed variable, a control, sample-size math, and a pre-registered success metric. Pulled by the `marketing` specialist. TRIGGER: "write A/B variants for the hero headline", "set up an A/B test on the pricing CTA", "draft multivariate copy for the signup page", "hypothesize and write variants for the email subject test", "run a copy test on the landing page above the fold", "design the test for the homepage hero". DO NOT TRIGGER for: writing the primary control copy itself (use `copywriter-long-form` / `copywriter-product-microcopy`), a single non-tested revision, positioning changes (use `marketing-positioning`), full launch plan (use `marketing-launch-plan`), platform-native social posts (use `social-media-platform-post`), a cold outreach sequence (use `sales-cold-outreach`).
Writes a campaign brief that commissions downstream work from copywriter, social-media, and seo — tight goal, audience, message, channels, assets-needed, deadlines, and success criteria. Pulled by the `marketing` specialist. TRIGGER: "write a campaign brief for <campaign>", "brief the copywriter and social-media team on the Q3 push", "draft the creative brief for the webinar promo", "commission a campaign against <theme>", "set up the campaign brief with goals and constraints", "brief the agency on the summer campaign". DO NOT TRIGGER for: overarching positioning (use `marketing-positioning`), full launch plan across phases (use `marketing-launch-plan`), writing the campaign assets themselves (copywriter / social-media / seo groups), A/B copy variant tests (use `marketing-ab-copy-variants`), sales outreach plan (use `sales-cold-outreach`).
Plans a product or feature launch across channels with a pre-registered metric stack, asset list, timeline (pre-launch, launch day, sustain), and per-channel owners. Pulled by the `marketing` specialist. TRIGGER: "plan the launch for <feature>", "draft the GTM launch plan", "what's our launch timeline and channel mix", "build the pre-launch → sustain plan", "coordinate the v2 launch across channels", "set up the launch success metrics". DO NOT TRIGGER for: positioning / ICP / category work (use `marketing-positioning`), a single campaign brief commissioning one piece (use `marketing-campaign-brief`), writing the announcement post itself (use `copywriter-long-form` or `social-media-platform-post`), release notes for engineers (use `technical-writer-release-notes`), sales enablement outreach (use `sales-cold-outreach`), A/B copy tests (use `marketing-ab-copy-variants`).
Matches all tools
Hooks run on every tool call, not just specific ones
Admin access level
Server config contains admin-level keywords
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
A Claude Code plugin that gives you self-evolving teams of specialist agents.
Guild turns a single coding session into a disciplined guild: /guild "<task>"
runs brainstorm, composes a team, writes per-specialist plans, assembles tight
context bundles, dispatches specialists, reviews, verifies, and reflects. Every
significant question becomes a structured decision. Every skill edit is a
versioned artifact with rollback. Nothing durable is written without passing a
gate.
agents/*.md per specialist.guild-principles), 13 meta
(the workflow spine + decisions + reflect + evolve + create-specialist +
rollback + audit), 3 knowledge (wiki ingest / query / lint), 5 fallback
(TDD, systematic-debug, worktrees, request-review, finish-branch — forked
from superpowers:* v5.0.7 under MIT, attribution preserved), 50 specialist
skills (2–5 per role)./guild, /guild:team, /guild:evolve, /guild:wiki,
/guild:rollback, /guild:stats, /guild:audit.SessionStart, UserPromptSubmit, PostToolUse,
SubagentStop, Stop, TaskCreated, TaskCompleted, TeammateIdle.scripts/evolve-loop.ts, flip-report.ts,
shadow-mode.ts, description-optimizer.ts, rollback-walker.ts,
trace-summarize.ts — plus scripts/agent-team-launcher.ts for the
opt-in tmux backend.mcp-servers/guild-memory/ (BM25 over the wiki
once it crosses ~200 pages) and mcp-servers/guild-telemetry/ (structured
trace query). Both stdio-only, no network. Guild runs without them.claude plugin marketplace add lookatitude/guild
claude plugin marketplace update guild
claude plugin install guild@guild
Restart Claude Code after installation so commands, skills, agents, hooks, and MCP servers are loaded into the next session.
Local development:
git clone https://github.com/lookatitude/guild.git
cd guild
claude plugin marketplace add .
claude plugin marketplace update guild
claude plugin install guild@guild --scope project
Restart Claude Code before running /guild from the project.
The agent-team backend is experimental. Enable only when teammates need to coordinate directly:
export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
guild-memory and guild-telemetry ship as stdio MCP servers under
mcp-servers/ but are optional — Guild works end-to-end without them.
Both ship pre-bundled (dist/index.js), so they cold-start under plain
node with no npm install step on first use.
Use guild-memory when the wiki crosses ~200 pages (ripgrep gets slow);
use guild-telemetry for structured trace queries over .guild/runs/.
/guild "Build a Stripe subscription flow, add tests, update the docs, draft a launch email."
The session will:
depends-on:..guild/context/<run-id>/.You confirm after brainstorm, team-compose, and plan. Post-plan runs with minimal interruption.
| Command | Purpose |
|---|---|
/guild [brief] | Full 7-step lifecycle: brainstorm → team-compose → plan → context-assemble → execute → review → verify |
/guild:team [propose|show|edit] | Manage the current team; edit --allow-larger lifts the 6-specialist cap |
/guild:evolve [skill] [--auto] | Run a skill through the evolve pipeline (paired evals → flip report → shadow mode → promotion gate) |
/guild:wiki [ingest <path>|query "..."|lint] | Wiki operations over .guild/raw/ and .guild/wiki/ |
/guild:rollback <skill> [n] | Walk a skill back n versions from .guild/skill-versions/ |
/guild:stats | Usage, success rates, flip counts, top-used skills, top-requested specialists |
/guild:audit | Security audit of installed scripts, hooks, permissions |
npx claudepluginhub lookatitude/guild --plugin guildUpstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Comprehensive startup business analysis with market sizing (TAM/SAM/SOM), financial modeling, team planning, and strategic research
v9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Complete creative writing suite with 10 specialized agents covering the full writing process: research gathering, character development, story architecture, world-building, dialogue coaching, editing/review, outlining, content strategy, believability auditing, and prose style/voice analysis. Includes genre-specific guides, templates, and quality checklists.
Comprehensive .NET development skills for modern C#, ASP.NET, MAUI, Blazor, Aspire, EF Core, Native AOT, testing, security, performance optimization, CI/CD, and cloud-native applications