Stress-test the user's ideas, plans, workflows, projects, and writing concepts through rigorous iterative challenge. Use when the user proposes a direction and wants deep pushback, dependency mapping, tradeoff analysis, alternatives, and a strength rating before deciding.

UK Companies House CLI — search companies, profiles, officers, filings, PSC, charges, insolvency, and agent-friendly JSON output aligned with rail-cli and tfl-cli. Use when looking up UK company records, directors, filing history, beneficial owners, charges, insolvency, or when an agent needs stable JSON envelopes with `ok`, `schemaVersion`, `command`, `requestedAt`, and nested `data.input` / `data.pagination`.

Deep-review a monorepo after a major refactor or rewrite by validating the branch, splitting API/web/runtime review lanes, and synthesizing a tight triage list.

Capture raw user dumps passed to the skill invocation into workspace-local markdown logs with no chat response.

UK fuel prices CLI — find nearby stations by postcode or coordinates, get station details, ranked by price/distance/freshness, and agent-friendly JSON envelopes with `--output <path>` for field projection. Use when checking UK petrol or diesel prices, finding cheapest fuel nearby, looking up station details by name or node ID, or when an agent needs structured fuel price data with data-quality advisories.

Capture screenshots and generate a tour GIF from a locally running web app for README, announcements, or social posts.

Curate and promote memory with signal-first judgment. Use when running memory checkpoints, consolidating daily notes, deciding what to store in durable memory, or when the user says “remember this”. Apply a strict rubric so only stable decisions/preferences/policies are promoted to long-term memory.

OpenClaw Blackbox CLI — inspect failed, stalled, expensive, or weird OpenClaw agent runs from local disk evidence; list sessions and failures; find the session behind a request; and generate terminal, Markdown, JSON, or HTML reports. Use when debugging OpenClaw cron runs, session failures, context overflows, missing transcripts/trajectories, message delivery issues, or when an agent needs deterministic run inspection without hidden LLM calls.

UK Parliament CLI — search bills, divisions, members, and written questions from official UK Parliament APIs, with stable JSON envelopes for agents. Use when looking up UK Parliament bills, members, votes/divisions, written questions, or when an agent needs structured Parliament API results.

Create a tight planning pack for a new software product before implementation — PRD, implementation plan, UI spec, source map, and repo-local AGENTS.md.

Build AI products where the prompt IS the product. Test-first prompt engineering, realistic lazy inputs, standalone test harness, iterate before building UI.

UK National Rail CLI — live departures, arrivals, station search, destination filtering, batch station search from stdin, and lightweight field selection for agents. Use when checking UK rail boards, resolving station names/CRS codes, filtering departures or arrivals, or when an agent needs machine-friendly station search via `rail search --stdin` or `rail search --select`.

Find context from past conversations, decisions, or discussions when current session memory falls short. Use when the user asks "did we talk about X?", "where did we decide Y?", or when you can't find something and are about to say "I don't know".

Unified review/research workflow for links, repos, posts, videos, and topics. Use when the user asks to review, research, summarize, sanity-check, or 'look into' something. Enforce safe defaults: never run untrusted repo code/tests/builds, prefer static analysis, write findings to a notes repo, return a concise TL;DR, and keep the main chat responsive via quick-ack + background execution for longer tasks.

Use when executing implementation plans with independent tasks. Dispatches fresh subagent per task with two-stage review (spec compliance then code quality).

Use when encountering any bug, test failure, or unexpected behavior. 4-phase root cause investigation — NO fixes without understanding the problem first.

London transport CLI — tube status, journey planning, live arrivals, disruptions, bike docks, and agent-friendly projection via `--output <path>` for route, arrivals, and bikes. Use when checking TfL line status, planning London journeys, getting live arrivals, finding bike docks, searching stops, or when an agent needs one value or subtree instead of the full response.

Create a user-centric weekly focus brief from workspace dumps and memory, with concrete actions and practical pattern analysis.

Create or refine agent skills with clear triggers, lean structure, and reusable references/scripts. Use when the user asks to create, rewrite, package, or improve a skill for recurring workflows.

Use when you have a spec or requirements for a multi-step task. Creates comprehensive implementation plans with bite-sized tasks, exact file paths, and complete code examples.

Draft, rewrite, humanize, and critique posts/emails/announcements in the user's authentic voice. Use when the user asks "write this in my voice", wants a public or private draft that sounds like them, wants to humanize AI-ish copy, or says "run the critic on this".