claude

workhorse

Agentic SDLC for forward-deployed engineering with risk-tiered human gates.

1 on GitHubby Muhammad Muhibullahv0.5.2MITupdated 1 month ago
/

Plugin workhorse from marketplace workhorse in muhibm1/workhorse-snapshot. Commands read from the repo's manifest — 1 month ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use and it starts local MCP servers. The exact commands are listed below; read them before installing.

What it installs

Skills (24)

  • approve

    Record your decision at a WorkHorse gate (Design = G2, Ship = G4, Deploy = G5 at tier 3). Approve advances the pipeline; reject re-enters the phase with your notes. Asks you about each decision the packet requested, writes approvals.md and commits it. Only you run this.

  • design

    Design phase only. Creates a WorkHorse change from "<problem> -> <outcome>" (or takes the active one), runs the designer, constraint auditor and eval designer, and presents the Design document (G2) without building anything. Replaces /workhorse:intent, /workhorse:spec and /workhorse:plan.

  • handover

    Handover phase for the whole engagement. Produces docs/handover so client engineers can own the system, regenerates the client CLAUDE.md, and optionally exports the WorkHorse workflow into the repo.

  • onboard

    Discover phase for a new client repository. Audits the codebase and business workflow, writes .workhorse/profile.yml, the codebase map, constraints, and a client CLAUDE.md, then presents G0 for you to confirm.

  • retro

    Run a WorkHorse retro on the last completed change, or triage an incident into a new change. Feeds lessons into CLAUDE.md, the profile, and evals as reviewable diffs.

  • review

    Review phase only for the active WorkHorse change. Runs the reviewers in parallel, sends fixable high and medium findings to the fixer and re-verifies, then the shipper writes ship.md, pushes the branch, opens the PR, and presents Ship (G4).

  • run

    Run the WorkHorse pipeline from "<problem> -> <outcome>" through to the next human touch (Design, Ship, or Deploy at tier 3), or resume the active change. Agents do the work; you approve at gates.

  • ship

    Deploy phase for the active WorkHorse change after Ship (G4) is approved. The shipper merges the PR and deploys to every environment the profile marks automatic; at tier 3 it writes release.md and presents Deploy (G5) for the human to run production.

  • status

    Show the active WorkHorse change, its tier, phase, gates by name (Design, Ship, Deploy), verification state, the clock against the tier budget, what is blocking, and optionally the engagement metrics.

  • sync

    Mirror WorkHorse changes into the client's Notion tracker so people who do not read git can see phase, tier, and which gates are waiting on a human. Run with --setup once per client to create the database.

  • verify

    Verify phase only for the active WorkHorse change. `wh.js verify` runs every profile check and writes the evidence; the verifier agent runs the automatable evals; the fixer loops with tests locked until green (max 5); verification.md is the record.

  • wh-adr

    Architecture decision record format for WorkHorse changes. Loaded by the designer. Use whenever a design choice has a plausible alternative a client engineer might later ask about.

  • wh-agent-rules

    Shared operating rules for every WorkHorse agent. Preloaded by all wh-* agents. Covers epistemic labelling, evidence before assertion, artifact discipline, and how to stop.

  • wh-compliance-financial

    Controls for WorkHorse changes in regulated financial services contexts (banking, lending, payments, trading, insurance) under regimes such as FCA/PRA, SEC/FINRA, DORA, SOX, and AML/KYC obligations. Loaded when profile.compliance.regimes includes financial.

  • wh-compliance-gdpr

    GDPR and UK GDPR controls for WorkHorse changes that process personal data of EU or UK residents. Loaded when profile.compliance.regimes includes gdpr. Use in spec, constraint audit, security review, and handover.

  • wh-compliance-hipaa

    HIPAA Security and Privacy Rule controls for WorkHorse changes that touch protected health information (PHI) in US healthcare contexts. Loaded when profile.compliance.regimes includes hipaa.

  • wh-compliance-pci-dss

    PCI DSS v4 controls for WorkHorse changes that touch card payments. Loaded when profile.compliance.regimes includes pci-dss. Use in spec, constraint audit, and security review of any payment path.

  • wh-compliance-soc2

    SOC 2 Trust Services Criteria evidence expectations for WorkHorse changes at clients undergoing or holding a SOC 2 report. Loaded when profile.compliance.regimes includes soc2 or iso27001. Use to make every change leave auditor-ready evidence.

  • wh-evals

    How WorkHorse turns "does it work" into measurable evidence. Loaded by the designer, eval designer, verifier, and incident triager. Use when designing acceptance cases, golden datasets, failure taxonomies, or turning an incident into a permanent test.

  • wh-handover

    What a WorkHorse engagement hands to a client so their engineers can own the system without the FDE. Loaded by the handover writer and adoption reviewer. Use when producing docs/handover or judging whether a change is maintainable by the client.

  • wh-language-standards

    Tells code-writing WorkHorse agents which vendored ECC language rules to read for this client's stack before writing or reviewing code. Preloaded by the builder, polish, and fixer.

  • wh-readable-code

    The readability and adoption standard for code WorkHorse produces in client repositories. Loaded by the designer, builder, polish, and adoption reviewer. Use when writing or reviewing any source file.

  • wh-review-packet

    How to write the WorkHorse Ship document (ship.md), the one document a human reads to approve Ship (G4), from templates/ship.md. Loaded by the shipper and the conductor. The Design document (brief.md) follows the same rules with its own template.

  • wh-security-baseline

    Security engineering rules and the pre-ship checklist applied to every WorkHorse change that touches auth, a database, RLS, functions, storage, secrets, CI, or an admin surface. Loaded by the designer, constraint auditor, security reviewer, builder, and shipper.

Subagents (28)

  • ecc-build-error-resolver

    Build and TypeScript error resolution specialist. Use PROACTIVELY when build fails or type errors occur. Fixes build/type errors only with minimal diffs, no architectural edits. Focuses on getting the build green quickly.

  • ecc-database-reviewer

    PostgreSQL database specialist for query optimization, schema design, security, and performance. Use PROACTIVELY when writing SQL, creating migrations, designing schemas, or troubleshooting database performance. Incorporates Supabase best practices.

  • ecc-fastapi-reviewer

    Reviews FastAPI applications for async correctness, dependency injection, Pydantic schemas, security, OpenAPI quality, testing, and production readiness.

  • ecc-pr-test-analyzer

    Review pull request test coverage quality and completeness, with emphasis on behavioral coverage and real bug prevention.

  • ecc-python-reviewer

    Expert Python code reviewer specializing in PEP 8 compliance, Pythonic idioms, type hints, security, and performance. Use for all Python code changes. MUST BE USED for Python projects.

  • ecc-rag-pipeline-reviewer

    Reviews RAG (Retrieval-Augmented Generation) pipelines for retrieval quality, chunking strategy, embedding choices, and evaluation coverage. Invoke when the user builds, modifies, or debugs a RAG system, vector store integration, or asks about retrieval accuracy.

  • ecc-react-reviewer

    Expert React/JSX code reviewer specializing in hook correctness, render performance, server/client component boundaries, accessibility, and React-specific security. Use for any change touching .tsx/.jsx files or React component logic. MUST BE USED for React projects.

  • ecc-silent-failure-hunter

    Review code for silent failures, swallowed errors, bad fallbacks, and missing error propagation.

  • ecc-typescript-reviewer

    Expert TypeScript/JavaScript code reviewer specializing in type safety, async correctness, Node/web security, and idiomatic patterns. Use for all TypeScript and JavaScript code changes. MUST BE USED for TypeScript/JavaScript projects.

  • wh-adoption-reviewer

    Review-phase agent. Judges whether a client engineer could maintain the change from the repository alone, scoring adoption 1 to 5 and listing concrete blockers in docs, naming, structure, tests, and operability. Use in parallel with the other reviewers.

  • wh-bug-reviewer

    Review-phase agent. Reads the change diff for logic errors, unhandled edge cases, race conditions, resource leaks, and incorrect error handling, reporting only findings it can state as a concrete failure scenario. Use in parallel with the other reviewers before the shipper writes the Ship document.

  • wh-builder

    Build-phase agent. Executes exactly one task from plan.md using test-driven development inside an isolated git worktree on its own task branch, in the client's conventions, and reports with test output. The conductor runs up to profile.build.max_parallel builders at once. Use only with a task number

  • wh-compliance-mapper

    Discover-phase agent. Given the client's industry, jurisdictions, and data inventory from the discovery analyst, decides which compliance regimes and controls apply, updates profile.compliance, and lists the controls each later phase must honour. Use during /workhorse:onboard or when the data invent

  • wh-conductor

    Drives a WorkHorse change through the six phases (design, build, verify, review, deploy, done), dispatching phase agents, keeping state.json current via the wh CLI, running Build as parallel waves of up to profile.build.max_parallel builders, and stopping only at Design (G2), Ship (G4) and, at tier

  • wh-conformance-reviewer

    Review-phase agent. Maps every requirement in spec.md to the code and test that implement it, and every change in the diff back to a requirement, flagging missing coverage and scope drift. Use in parallel with the other reviewers.

  • wh-constraint-auditor

    Design-phase agent. Audits spec.md, plan.md and brief.md against the security baseline, the client's compliance regimes, privacy defaults, and the profile's constraints before any code exists. Writes findings by severity into the spec's constraint audit table; only a high finding sends the designer

  • wh-designer

    Design-phase agent. In one session turns "<problem> -> <outcome>" into the risk tier, spec.md with ADRs, evals.md within the tier's case limit, plan.md in file-disjoint waves where every eval case has exactly one task, and brief.md, the Design document the human approves at G2. Never stops to ask; e

  • wh-discovery-analyst

    Discover-phase agent. Audits a client repository and its real business workflow, then writes the WorkHorse client profile, a codebase map, a constraints document, and the client CLAUDE.md. Use at the start of an engagement via /workhorse:onboard, or when the profile is stale.

  • wh-eval-designer

    Design-phase agent, tier 2 and above only. Reviews and extends the evals.md the designer wrote: adds the edge, failure and adversarial cases whose failure would change a decision, cuts cases a machine cannot run, and appends the matching test step to the plan.md task that owns each added case, withi

  • wh-fixer

    Verify and review-phase agent. In verify mode fixes the specific failures the verifier reported, changing implementation code only while the test-lock hook is active; in review mode, run with the lock off, fixes every high or medium reviewer finding that came with a concrete fix, in implementation o

  • wh-handover-writer

    Handover-phase agent. Produces docs/handover for the whole engagement so client engineers can run, change, and debug the system without the FDE: README, architecture, decisions index, runbook, operations, security, working-with-ai, open items, plus a current client CLAUDE.md. Use via /workhorse:hand

  • wh-incident-triager

    Maintain-phase agent, designed to run headless from monitoring or on a schedule. Given an alert, incident, scan result, or failing metric, diagnoses read-only, writes a reproducing eval case, and writes the problem and outcome for a new change to enter the pipeline through /workhorse:run and its nor

  • wh-polish

    Build-phase agent. One pass over the whole change's diff after the last build wave, for naming, structure, comments, client conventions, dead code, needless abstraction, duplicated logic, over-general helpers, and defensive code for impossible cases. Fixes what is safe, lists the rest, keeps tests g

  • wh-retro-learner

    Done-phase agent, run in the background after `WH phase done` with the change id given explicitly. Writes retro.md with the clock judged against the tier budget, records lessons as instinct files, proposes edits to the client CLAUDE.md "Mistakes to avoid", the profile, and evals, and commits retro.m

  • wh-security-reviewer

    Review-phase agent. Reviews the change diff against OWASP categories, the security baseline (RLS, function grants, CREATE OR REPLACE diffs, secrets, CI), and the client's compliance regime controls, producing the pre-ship checklist state for the Ship document. Mandatory at tier 2 and above. Use in p

  • wh-shipper

    Review and deploy-phase agent. In ship mode, after the reviewers and the post-review fix, writes ship.md (the Ship document the human reads at G4), pushes the change branch, and opens the PR. In deploy mode, after G4 approval, merges the PR, deploys to every environment the profile marks automatic,

  • wh-tracker-sync

    Mirrors WorkHorse changes into the client's Notion tracker so people who do not read git can see what was asked, what phase it is in, and which gates are waiting on a human. One direction only, repo to tracker. Never blocks the pipeline. Use from /workhorse:sync or from the conductor after a phase o

  • wh-verifier

    Verify-phase agent for the evals. `wh.js verify` has already run the profile's checks and written their evidence; this agent runs every automatable case in evals.md, completes verification.md (Evals table, the comparison sentence, known-failure citations, notes) and records the final status. Cannot

MCP servers (1)

  • studbook

    runs: node ${CLAUDE_PLUGIN_ROOT}/scripts/studbook-mcp.js

Hooks (11)

  • SessionStart

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/session-context.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/protect-paths.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/test-lock.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/worktree-fence.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/gate-guard.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/bash-guard.js"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/gate-guard.js"

  • PostToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/format-on-write.js"

  • PostToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/artifact-check.js"

  • Stop

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/verify-before-stop.js"

  • SubagentStop

    node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/verify-before-stop.js"

View source on GitHub

Similar plugins

Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).

39 skills

mattpocock/skills 323,108

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

claude-mem

claudecodex

Persistent memory for Claude Code: the next session starts where the last one left off.

29 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

mem0

claudecodex

Cross-session memory and token savings for coding agents.

6 skills · 1 agent · 1 MCP server · hooks

mem0ai/mem0 68,893

context7

claudecodex

Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.

1 MCP server

upstash/context7 64,709

Automates browser interactions for web testing, form filling, screenshots, and data extraction

12 skills

vercel-labs/agent-browser 50,373

Maintain workhorse?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex pluginslisted