mattpocock-skills
claudeMatt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Build agentic data apps with ORM models, deterministic checks, and bounded development workflows. Requires the get-nick-done Python package.
Plugin gnd from marketplace gnd-marketplace in vasiloglou/gnd-public. Commands read from the repo's manifest — 2 days ago.
componentsThe project's component index (docs/components/), which says what already exists in the project and its dependencies. Consult mode searches it by capability (scripts/components_index.py search) and records a reuse / adapt / reject decision. Update mode edits entries, assigns new modules, marks dupli
plan-review-panelCompatibility entry point for an explicitly selected advisory plan review using the maintained GND panel; no inherited vendor runner.
process-milestoneExecute or resume a named development milestone through its phases, required reviews and closure evidence. Supports auto=true for explicitly delegated milestone approval gates. Use for whole-milestone execution; use process-phase for a single phase and start-milestone for planning only.
process-phaseExecutes one phase of a milestone plan under that milestone's own protocol -- harvest obligations, author the spec and stop for its owner gate, implement and verify, review the diff and refactoring, then record and commit. Use for requests to run, process, execute, specify, review, or close a number
start-milestonePrepare or resume a development milestone plan, turning requirements and carryovers into phased deliverables, measurable gates and the first draft phase spec. Use when asked to start, plan or initialize a repository milestone. Does not execute phases or close a milestone.
gnd-adoptMigrates a GSD project into GND: recovers an ORM model from the codebase via gnd-model-from-code verbatim, maps every GSD REQUIREMENTS id into the model as digest-anchored imported provenance or an explicit open-question / superseded / history disposition, verifies the whole mapping with the bundled
gnd-codegenTurns an ORM model into a runnable plain-Python agent project with `gnd codegen MODEL --out DIR`, then writes the EXTENSION half — the architecture declaration, domain notes, and dependency list — that regeneration is guaranteed never to overwrite. Use when asked to "generate an agent from a model",
gnd-conformanceChecks generated-project agreement with `gnd conformance MODEL DIR` and records the evidence a codegen experiment needs before close. Use for model-code drift checks, conformance runs, or committed conformance evidence. Four assertions cover recorded derived-file digests, model currency, the constru
gnd-dev-sessionDrives GND's data-driven development session — `gnd exp init` to declare the questions, a versioned session contract binding inputs and a verifier plus exact outputs, constraint floors, sealed tests, or a transition walk, `gnd exp run` to invoke a bounded legacy or locally declared adapter, and `gnd
gnd-exportExports and imports pinned Apache Ossie ontologies with loss reports. Exports finite Lean models with unproved obligations, finite grounded witnesses, or explicitly selected bounded standalone Souffle derivation programs. Native compilation/execution and positive proof replay are separate optional c
gnd-extractExtracts an ORM population from a natural-language document — you read the document, write extracted.population.json against the model's shape-only projection, write the digest-bound extraction sidecar beside it, and run `gnd verify` over your own output. Use when asked to "extract a population from
gnd-judge-fact-typesJudges whether the groups the `gnd discover --families fact-type` family formed duplicate fact types the model already has, from one type-evidence file and the model only, and writes the `duplicates` judgements file `gnd discover --duplicate-judgements` reads back. Use when asked to "judge the induc
gnd-model-evolveSafely evolves an existing valid GND ORM model through explicit add, modify, rename, delete, or include re-pin operations, complete digest-bound blast-radius review, model CAS publication, live linked-project conformance, and idempotent finalization. Use when asked to change or migrate model.orm.yam
gnd-model-from-codeRecovers a reviewable ORM 2 model from an existing codebase through one bounded, static pass: inventory explicit source/test/docs/config/schema roots, retain evidence and ambiguity, author model.orm.yaml plus its complete verbalization and recovery report, then run gnd recovery verify. Use for brown
gnd-model-from-textBuilds an ORM 2 conceptual model (model.orm.yaml with full constraint provenance — explicit / sample / common-sense / llm-inferred / user-supplied, alethic vs deontic modality on every constraint) from a natural-language domain description by walking the user through Halpin's Conceptual Schema Desig
gnd-model-multivaluedDecides how to model the multi-valued groups that the `gnd discover --families fact-type` family counts but does not propose, from one type-evidence file and the model only, and writes the `shapes` judgements file `gnd discover --shape-judgements` reads back. A variant of the `rai-orm-from-schema` d
gnd-new-projectOpens a GND project in a repository by scaffolding the .gnd-planning directory — the one place all GND planning artifacts live — with its PROJECT.md one-pager and the first milestone directory, then makes the opening commit. Use when asked to "start a GND project", "gnd new project", "open a milesto
gnd-orm-coreThe ORM-YAML and population format reference library for GND — the single source of truth for the model.orm.yaml spec, constraint vocabulary, provenance/modality policy, verbalization patterns, and population interchange format. Consult when authoring, validating, or checking GND models, or when any
gnd-render-docRenders a natural-language document from an ORM population plus the model's constraint-blind verbalization, plants the three mandatory distractor kinds, writes the digest-bound document manifest, and runs `gnd doccheck` until it exits 0. Use when asked to "render a synthetic document from a populati
gnd-review-panel-lightRuns the same bounded six-reviewer advisory workflow as gnd-review-panel, but uses a lighter supported per-run effort: high by default or medium when the user explicitly selects it. Use for faster or lower-cost multi-model review evidence over one plan, one experiment outcome, or one spec/results/de
gnd-review-panelRuns GND's bounded six-reviewer advisory panel over one plan, one experiment outcome, or one spec/results/delta bundle with `gnd review panel`, then validates and renders a PANEL-REVIEW.md with `gnd review finalize`. Use for multi-model review evidence over one of those input classes. Six fixed revi
gnd-solveA process guide for the full nine-station GND loop. The `gnd solve` harness owns the ordered validate, synth, codegen, exp-run, and conformance command walk; this skill points to that command's help while retaining the routes and stable station numbers for opening a project, authoring or recovering
gnd-update-review-panelUpdates the five terminal-agent harnesses used by GND's active review panel (Claude Code, Codex CLI, Qwen Code, Gemini CLI, and Kimi Code), enforces Qwen's OpenRouter route, then refreshes and validates the fixed reviewer model and prompt-transport policy from current official provider documentation
Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
5 skills · hooks
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
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.