mattpocock-skills
claudeMatt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Shode House 4.1.0: 6 agent types, 20 skills, durable workflow and independent review.
Plugin shode-house from marketplace shode-house in shode666/shode-house. Commands read from the repo's manifest — 12 days ago.
domain-coreShared contract for domain expert personas, requiring an AI persona disclaimer and primary-source citations for any claim about a regulation, standard or protocol, and explicit marking when a source cannot be verified. It governs domain claims, not general project evidence or which expert is needed.
review-checklistCoordinate independent review of an implemented change by code, runtime, spec, security and domain with severity and merge gates, not self-approval.
shode-house-deliverableOutput and completion contract for anyone producing a deliverable - durable artifact, re-checkable evidence for every claim, no placeholders, no completion claim without real tool output. Not how work is routed or reviewed.
shode-house-disciplineBaseline conduct for team members: evidence, scope, language and handoff. Not phase order, ownership or the safety floor (see ask).
shode-house-routingDecide who owns a request, covering type of work, responsible role, conflict resolution, parallel versus sequential execution, and whether a domain specialist is needed because the work requires a domain rule decision rather than merely using domain vocabulary. Phase order and approvals are not deci
shode-house-workflowCoordinate multi-phase delivery with phase gates, recorded approvals, run recovery and drift detection, governing how work proceeds, not who owns it.
drainBatch-drain ready tickets with isolated workers, independent review, serial merge and evidenced closure (จัดงานที่พร้อม, ปิดงานค้าง). For work not split into concrete, independent tickets, use decompose.
incidentRespond to active production impact such as an outage, degradation, data loss or breach by assigning severity and mitigating first, then a blameless postmortem. Not for bugs not currently hurting production users, nor planned maintenance.
secureThreat model and review a design or change that crosses a meaningful security boundary, such as authentication, authorization, secrets, cryptography, personal or payment data, untrusted input or public exposure. Not for a live incident or unrelated local refactors.
sloDefine reliability targets for a production service, covering service level indicators and objectives, error budget policy and burn-rate alerting. Not for responding to a live outage, nor for internal tools or one-off batch jobs.
cavemanSwitch to an ultra-compressed reply style that drops filler but keeps technical terms, code, numbers and security warnings verbatim, only when the user explicitly asks for terse output. It is never the default and ends when the user asks for normal prose.
ui-testAutomated browser-level end-to-end, visual regression and accessibility testing of a changed, rendered user interface, plus the UI quality rules every frontend change follows. Not for backend-only changes, unrendered designs, or site performance audits.
web-qAudit and improve a public-facing website across Core Web Vitals, performance budgets, SEO, structured data and security headers against agreed measured thresholds. Not for internal or API-only services, nor for functional or accessibility testing of one changed screen.
api-contractProtect compatibility when changing a shared or public interface, such as a REST, GraphQL, gRPC, event or SDK contract with consumers outside the deploy unit, through breaking-change classification, versioning, deprecation and contract tests. Not for private function signatures or internal database
askEntry point for engaging the Shode House software team on consultation, design, implementation, review or resuming an engagement. Asking to explain or review does not authorize edits.
automate-testSet up a project's test automation strategy and infrastructure, meaning the mix of unit, integration and end-to-end tests, CI gates and thresholds. Not for the tests of a single change, one changed screen, or spikes and one-off scripts.
data-migrationPlan and carry out a schema or data migration, such as DDL, backfill, dual write or expand-contract rollout, with rehearsed rollback and extra care for money and regulated data. Not for ordinary SQL queries, disposable local databases or a manual one-row production fix.
decomposeSplit an agreed, signed-off scope into small, independently verifiable outcome slices with owners, acceptance criteria and explicit dependencies. Not for work whose finished shape is still unclear, tasks already small, or a backlog already verified and ready.
dev-gateApply test-first development and quality gates during implementation or refactor of production code, ending with validation evidence before handoff. Not for spikes, generated code, pure configuration changes, or investigating a problem whose cause is still unknown.
diagnoseDebug an error, crash, failing test, wrong output, regression or slowdown (พัง, ไม่ทำงาน, ช้า, ทำไมถึง, bug) by reproducing and isolating the root cause before fixing. For live production impact, use incident.
/shode-house:askWork with the full Shode House team.
/shode-house:consult[shode-house] ปรึกษาด่วน — route ไปหา agent ที่เหมาะที่สุด (ไม่รัน pipeline เต็ม)
/shode-house:design-system[shode-house] Smart Spec pipeline (Phase 1a `plan` (requirements mode) + `plan` (architecture mode) parallel; Phase 1b `design` + domain conditional). Flags: --stop = หยุดที่ spec ไม่ suggest implement (proposal mode); --estimate = เพิ่ม T-shirt sizing step
/shode-house:implement[shode-house] Implement (`build`) + UI Check (`design`) + review axes + Triage — Smart Coop Phase 2-4
/shode-house:init[shode-house] Init — scaffold project ใหม่. Default: interactive wizard (`operate` (deploy mode) + `plan` (requirements mode) + tracker). `--quick <stack>`: direct `operate` (deploy mode) Docker-first (replaces /setup-project)
/shode-house:review[shode-house] Code review + security (review axes: standards, spec, runtime, security, domain, ui) — รับ path, Jira ID, หรือ bug description (+ screenshot)
buildImplements authorized features and fixes across the project stack (frontend, backend, business logic, databases, integration) with behavior tests; staff-grade briefs cover cross-team consistency, tech radar and refactor strategy. Owns implementation evidence, not independent acceptance.
designUX research, information architecture, wireframes and design tokens, design system, accessibility (WCAG 2.2 AA), the ui review axis (Phase 3a), and design-run requests. Design authority; has no Bash and never writes product code.
operateRuns delivery and reliability - Docker-first setup, CI/CD, IaC, deploy builds (Phase 5), and SLO, error budget, incident, runbook and postmortem work (Phase 6). Deploys only what is authorized.
planProduct discovery, requirements and acceptance criteria, architecture and ADR, the spec review axis, and domain consult or the domain review axis (a domain reference loaded). Never writes product code.
secureThreat modeling (STRIDE/LINDDUN, Phase 1c), the security review axis, SAST/DAST, secrets management, pen testing and security headers. Sole owner of security depth; does not yield to "low risk".
verifyIndependently verifies a change on one axis per spawn - standards (7-dimension code review, unit-test quality) or runtime (integration, contracts, end-to-end, load, accessibility automation). Requirement conformity is the plan spec axis; deep security is secure.
Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
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
Automates browser interactions for web testing, form filling, screenshots, and data extraction
12 skills
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams
6 skills · 7 commands · 4 agents
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.