cleanexpo
- 430 skills
- 0 followers
- last month last updated
- ▌ Design System · cleanexpoDesign stack orchestrator for Pi-CEO. Routes UI work to the correct specialist skill (design-intelligence, ui-component-builder, impeccable, visual-qa). Start here for any design task.
- ▌ Intent Parser · cleanexpo bundleClassify an inbound Telegram message into one of six intent types so the Chief of Staff can route to the right role. Use this on every Telegram message that lands in Pi-CEO before any routing decision.
- ▌ Launch Review · cleanexpoAggregate the repo's existing audit skills into ONE prioritized launch-readiness report through four lenses — PM journey, growth/SEO, engineering, design — adding only the two lenses nothing else covers (end-to-end PM journey + growth/SEO conversion) and fanning out to security-audit, impeccable, agentic-review, and leverage-audit for the rest. Review only; never fixes. Use on "is this ready to launch" or as the review step of /ship-it.
- ▌ Margot Bridge · cleanexpoBridge from Pi-CEO orchestrator to Margot's standalone Gemini-powered research MCP server at ~/.margot/. Use when intent is research-shaped (deep_research / deep_research_max) or image-gen, or when corpus diagnostic is needed. Margot is owned separately; this skill never modifies her source.
- ▌ Tao Boomerang · cleanexpoToken-efficient one-shot dispatch with summary-only return. Caller asks a question, the worker replies with ONLY the answer — no preamble, no chain-of-thought, no markdown unless the answer requires it. Composes on `_run_claude_via_sdk`. Use when you need the answer, not the conversation.
- ▌ UI UX Pro Max · cleanexpoFull-stack design workflow skill. Orchestrates all four design layers (intelligence, build, audit, visual-qa) end-to-end with maximum quality settings. Use for ambitious UI work — new pages, design system extensions, and design sprints.
- ▌ Agent Workflow · cleanexpoUse when defining, scoping, promoting, or debugging an agent or ADW — the 5-part agent contract, scoping formula, promotion ladder, and 4-way break diagnostic.
- ▌
- ▌
- ▌ Chrome Browser · cleanexpoUse when the user asks to browse, click, navigate, fill a form, screenshot, or read content from a live website via the claude-in-chrome MCP tools — or whenever a task needs to drive an actual Chrome tab rather than fetch static HTML.
- ▌ Claim Verifier · cleanexpoThe runnable cite-or-cut gate for FINISHED Nexus copy — walks every checkable factual claim in a completed asset, resolves each to a substantiation record or cuts it, runs the Australian Consumer Law excluded-tactics deny-list, and freezes the Substantiation Ledger before publish. Use as the last gate before any copy asset ships — landing page, service page, ad, email, franchise-recruitment copy. Triggers on "cite-or-cut", "substantiate this copy", "claim check", "can we publish this", "freeze the ledger", "is every claim proven", "ACL check the copy". NOT research (that is `storm`), NOT trust scoring (that is `eeat`).
- ▌ Claude Runtime · cleanexpoRules for invoking Claude correctly — subprocess vs SDK mode, stream-json parsing, MCP SDK imports, Railway vs local environment differences.
- ▌ Control Design · cleanexpoUse BEFORE writing or changing any check, test, gate, guard, assertion, exclusion, fixture or startup config branch — while the control is being built, not when its result is read. Catches controls that cannot fail, exclusions whose premise has expired, checks that know a fixed set and go stale, verifications that require handling the thing they protect, and missing-config branches that invent a credential instead of refusing to start. Triggers on "add a check", "write a test for", "gate this", "skip/exclude this path", "assert that", "guard against", or any first run of a new control.
- ▌ Design Approve · cleanexpoLabelling + cache step. Once the user approves a variant in design-iterate, this skill writes the canonical brand spec files (`{slug}.design.md`, `{slug}.motion.md`, optional `{slug}.scene.md`), generates the standalone `{slug}.html` preview file, snapshots the iteration to the approved-variant cache, and triggers `remotion-brand-codify` to write the BrandConfig.ts runtime fields. Treats approval as a commit point.
- ▌ Design Iterate · cleanexpoIteration loop driver for brand design. Calls design-board to generate N variants, renders each at three breakpoints (mobile/tablet/desktop) via the preview-canvas + Claude Preview MCP, presents screenshots to the client (the user) via AskUserQuestion, captures feedback, refines, and repeats until the user approves a variant. The visible canvas the client uses to actually SEE designs.
- ▌ Email Listener · cleanexpoConvert inbound Gmail messages into Pi-CEO sessions via webhook → intent-parser → orchestrator. Closes the "no email-triggered session creation" gap. Use this when Phill wants Pi-CEO to act on emails — replies, ticket creation, research dispatch — without him CC-ing it manually.
- ▌ Launch Charter · cleanexpoThe immutable governance charter every autonomous launch agent loads FIRST. Defines what agents may improve freely vs. what they may never touch (permissions, spend caps, safety, irreversible actions). Binds the generic self-improvement rails to Pi-Dev-Ops's real AGENTS.md boundary matrix. Use at the top of /ship-it, every cron run, and before any builder agent starts.
- ▌ Leverage Audit · cleanexpo12 Leverage Points diagnostic for agent autonomy — with explicit scoring rubrics and ROI guidance.
- ▌ Margot Routing · cleanexpoLinear routing checklist for Margot — place broad ideas in the parent project first, then copy to vertical projects only where scope belongs. Use when Margot creates tickets, triages founder ideas, or spans portfolio repos.
- ▌
- ▌ Pi SEO Scanner · cleanexpoScan interpretation specialist for Pi-SEO findings. Applies blast-radius scoring to prioritise findings, detects false positives, classifies into fix-now/schedule/suppress/investigate buckets, and recommends scan config changes per project.
- ▌
- ▌ Security Audit · cleanexpoSenior Security Officer (15+ years white/black hat experience). Deep-audits codebases for OWASP Top 10, supply-chain risks, secrets exposure, auth flaws, injection vectors, CSP misconfigs, and weak crypto. Produces a prioritised CVE-style finding report with CVSS scores and remediation steps.
- ▌ Senior Harness · cleanexpo bundleUse before substantive project work that requires governed discovery, delegation, verification, or recovery from repeated failure.
- ▌
- ▌
- ▌ Tier Evaluator · cleanexpoQA agent that grades output against explicit acceptance criteria. Skeptical by default. Blocks on failure.
- ▌
- ▌ Autonomy Ladder · cleanexpoCapability-tiered autonomy gating for all agent action — maps the DeepMind AGI→ASI continuum (arXiv:2606.12683) onto four autonomy tiers (L0 advise → L3 strategic/irreversible), each with an explicit gate. Use to decide "can this agent do this action un-gated, or must it stop for a human/Board?" before any autonomous or multi-move execution, and to design where the gate is ENFORCED. Triggers on "autonomy ladder", "autonomy gate", "can the agent do X autonomously", "gate this action", "should this escalate", multi-move executor design. The formal, capability-scaled version of the decision-rights matrix.
- ▌ Browser Routing · cleanexpoUse when a task needs a browser — automating a web flow, reviewing or improving UI, debugging a page, recording evidence, or QA-testing a deploy — before picking a browser tool or writing a browser-touching skill.
- ▌ Context Cockpit · cleanexpoAudit and reclaim a session's context/token budget in one pass — read /context, trim MCP tools, clear vs compact, CLI-over-MCP, model ladder, red-MCP triage.
- ▌ Cross Tool Flow · cleanexpoDeclarative YAML/JSON authoring surface over dispatcher-core. Lets the user (or Margot, or Scribe) define multi-step cross-tool flows in plain text — chain Linear, Gmail, Calendar, Margot, Composio MCP tools — and run them via Dispatcher. Wave 2 surface; engine is dispatcher-core in Wave 1.
- ▌ Dispatcher Core · cleanexpo bundleCross-tool workflow primitive. Executes a declared sequence of steps that chain Linear, Gmail, Calendar, Margot, Composio, or any registered MCP tool. State persisted to .harness/dispatcher_state.json so kill-switch + crashes don't lose progress.
- ▌ Forward Planner · cleanexpo bundleResearch a project and plan 15+ moves ahead before building anything. Establishes the win condition (Definition of Done), reads the board (repo + PORTFOLIO registry + live external research), computes the gap, lays a roadmap spine of 15+ verifiable moves, forks branch points where outcomes are contingent, and red-teams the end-state to surface the "move-16 surprise" before it bites. Produces BOTH a human-readable foresight brief AND a machine-readable structured plan the swarm can execute. Use this FIRST whenever you (or the autonomous loop) are about to build, scope, or roadmap a project — and whenever someone asks "what's the full path to X", "what are we missing", "plan N moves ahead", "think this through", or talks about strategy, foresight, lookahead, sequencing, dependencies, or roadmaps. Use it even when the request just says "build project X" — map the whole path before executing the first step.
- ▌
- ▌ Incident Memory · cleanexpoWrite symptom, cause, fix, reviewer verdict and outcome to a searchable store — successes and failures alike. This is the store `classify` reads to decide known versus novel. Use as the final step of every self-healing chain, including chains that stopped early or gave up.
- ▌ Piter Framework · cleanexpo5-pillar AFK agent setup - Prompt, Intent, Trigger, Environment, Review.
- ▌ Product Manager · cleanexpoSenior Product Manager Engineer (15+ years SaaS delivery). Audits feature completeness vs advertised capabilities, identifies user journey gaps, evaluates documentation quality, and produces a prioritised product backlog aligned with business outcomes.
- ▌ Remotion Script · cleanexpoUse when a Remotion marketing video needs a timed script, scene narration, on-screen text, CTA, and voice pacing that fits a single ElevenLabs voice.
- ▌ Scheduled Tasks · cleanexpoGuidelines for writing reliable scheduled task prompts via the Claude scheduled-tasks MCP.
- ▌ Session Handoff · cleanexpoDurable session handoff (/session-handoff). Gates the tree green via scripts/handoff-loop.sh, classifies the session SHIPPED / READY-TO-SHIP / WIP-BLOCKED against a Definition-of-Done, then generates a precise handoff before stopping, switching terminals, opening a PR, handing work to another agent, or resuming later. Refuses to present unfinished work as a clean stop; for ready work hands off the exact merge-gate-guarded ship command. Writes a durable report + healthcheck log; the "1" of the 1-2 combo with /resume-from-handoff. Captures what was done, decisions locked, what shipped, key files, running state, verification, deferred/open questions, exact pickup point, risk notes, and a quality check.
- ▌ Tao Context Vcc · cleanexpoDeterministic, LLM-free conversation compactor for TAO sessions. Port of @sting8k/pi-vcc. Reduces transcript bytes via tool-output dedup, verbose-block head/tail truncation, repeat-pattern collapse, and whitespace normalisation. No model calls — runs in the hot path with zero token cost.
- ▌ Brand Ambassador · cleanexpoGenerate brand-consistent content, copy, and messaging aligned with Pi-CEO tone and values.
- ▌ Calendar Watcher · cleanexpoConvert calendar event creation/modification/deletion into Pi-CEO actions. Use for time-anchored triggers — meeting prep 30min before, post-meeting summary, conflict detection, schedule reflow when an urgent block lands. Pairs with email-listener for the "calendar invite arrived in email" pattern.
- ▌ Geo Optimization · cleanexpoGenerative + Answer Engine Optimization (GEO/AEO) standard for AI search visibility — the 2026 successor to SEO. Use when adding or auditing GEO infrastructure (llms.txt, JSON-LD schema, WebMCP, Lighthouse Agentic) OR AEO content-layer work (becoming the cited/quoted answer inside ChatGPT, Perplexity, Gemini, Google AI Overviews) on any portfolio site or via the Synthex client-audit module.
- ▌ Northstar Shipit · cleanexpoNorthStar add-on for /northstar and /ship-it. Removes noisy branches, picks the highest-value straight path to ShipIt, and hands only the approved launch-ready lane into the existing launch-charter → ship-it → ship-chain machinery.
- ▌ Pi Ceo Docparser · cleanexpoParse PDF / DOCX / TXT into structured ParsedDoc (text + pages + tables + metadata). Pure deterministic extraction, no LLM call. Foundation for ICP-research workflows; consumed by marketing-icp-research.
- ▌ Proof Discipline · cleanexpoUse whenever about to mark a test, gate, verification or task as passing / GREEN / complete / fixed / shipped — the moment a CLAIM is made, not when a check is written. Catches vacuous verification, claims asserted but never checked against the live system, silent caps, deployed-vs-source drift, coverage gaps that hide non-functioning code, and green results whose control was aimed at the wrong target. Triggers on "tests pass", "it's green", "verified", "done", "complete", "fixed", "shipped".
- ▌ Remotion Editing · cleanexpoUse when a Remotion video has timing, sync, transition, caption, pacing, or audio-fit risk and needs editing discipline before render.
- ▌ Tao Context Mode · cleanexpoSummary-index + on-demand expansion. Walks a repo once, summarises every source file (~200B vs raw 5-50KB), and serves the FULL file only when callers `expand(path)`. Port of `context-mode`. No LLM calls — pure regex + first-comment synopsis. Deterministic; sha256-keyed for invalidation.
- ▌ Tao Tdd Pipeline · cleanexpoTest-first iteration pipeline. Composes on tao-loop + tao-judge to enforce red→green→refactor discipline. The judge is bound to "all tests pass + new test files modified". Without test files in the diff, GOAL_MET is rejected.
- ▌ Creative Director · cleanexpo bundleUse when someone asks to build, design, make, create, or generate any deliverable from a short prompt — a deck, web page, report, poster, campaign, brand look, prototype — especially when the prompt is underspecified. Resolves the subject and client persona, grills only the gaps with one checkbox screen, routes to the right specialist skills and MCPs, and runs a polish gate before handoff. Do not start generating for such requests without this skill.
- ▌ Output Tournament · cleanexpoUse when a creative brief has multiple plausible directions and picking wrong is costly — landing-page hero copy, taglines, headlines, hooks, video concepts, design directions. Generate several diverging variants, judge them independently, synthesise a winner. Not for high-stakes decisions or cross-domain review of one shared draft.
- ▌ Prove The Failure · cleanexpoBuild a tight, red-capable feedback loop that drives the actual failure path and asserts the exact symptom, before any hypothesis is formed. One command, already run once, deterministic, fast, agent-runnable. Use at the start of any self-healing chain, or whenever something is broken/failing/slow and you are tempted to read the code and guess. First link in the self-healing chain; hands off to `contain`.
- ▌ Remotion Designer · cleanexpoVisual-design specialist for Remotion compositions. Sets layout grid, typography hierarchy, white-space, scene framing, and visual rhythm for a given composition + brand. Triggered after remotion-composition-builder for QA, or upfront when authoring a new composition type. Produces a layout spec consumed by remotion-composition-builder.
- ▌ Sandcastle Runner · cleanexpoThe primitive. Wraps a single Sandcastle invocation (`npx sandcastle run`) so Pi-CEO orchestrator can spawn AFK code-executing agents in isolated Docker / Podman / Vercel-Firecracker / Daytona sandboxes. Pluggable agent (Claude Code / Codex / Pi / OpenCode) and pluggable sandbox provider. Stream stdout into session.logs. Pre-flight kill-switch + concurrency cap. Closes Wave 5
- ▌ Skybridge Rollout · cleanexpoUnite-Group rollout plan for Skybridge MCP-Apps framework — which businesses get an MCP App, in what order, with what auth, and how to productize it via Synthex. Invoke when planning, scoping, or building any MCP App across the portfolio.
- ▌
- ▌ Tier Orchestrator · cleanexpoThe orchestrator plans, decomposes briefs, and delegates to lower tiers.
- ▌ Vercel Env Puller · cleanexpoPer-project Vercel env-var manifests (NAMES + targets only — never values). At Sandcastle launch time, resolves the named vars in-memory via `vercel env pull`, injects into the sandbox via Sandcastle's `agent.env`+`sandbox.env`, and never writes a secret to disk inside Pi-CEO. Closes Wave 5
- ▌ Adversarial Review · cleanexpoReview a proposed fix with a different model in a fresh session that has no access to the builder's reasoning. Two axes, standards and spec, run as independent sub-agents. It flags; it never fixes. A fail returns to `propose-fix`. Use after `propose-fix`, before anything is applied.
- ▌
- ▌
- ▌
- ▌ Design Canvas HTML · cleanexpoGenerates a self-contained `{slug}.html` preview file that demonstrates a brand's full design system (palette + WCAG contrast + typography scale + component recipes + motion samples + optional 3D scene) inline — no server, no build, opens directly in any browser. Used by design-approve to ship a portable canvas alongside the canonical spec files. Mirrors the routes in preview-canvas/app/brand/[slug] but flattened into one HTML file.
- ▌ Pi SEO Remediation · cleanexpoRemediation advisor for Pi-SEO findings that cannot be auto-fixed. Produces per-finding remediation cards with tier classification, step-by-step fix instructions, before/after code examples, verification commands, and effort estimates.
- ▌
- ▌ Specialist Council · cleanexpoAI Communication Intelligence layer — the bidirectional cross-specialist consultation protocol that makes specialist agents TALK TO EACH OTHER before an artifact is finalized. Use when a task spans domains and the output needs input from multiple specialists (e.g. marketing copy that must satisfy SEO + GEO + E-E-A-T + brand voice, or a video that needs Remotion + HeyGen + design + brand). Any orchestrator or /spm invokes it to pull structured input from the RIGHT specialists and fold it back in. Triggers on "AI communication intelligence", "agents talking to each other", "cross-specialist", "specialists should consult each other", "get input from all relevant specialists", "council". NOT a builder and NOT a replacement for the single-domain orchestrators — it is the round-trip consultation they lack.
- ▌ Design Intelligence · cleanexpoMaster design context skill. Reads/writes DESIGN.md, references 66 brand archetypes from getdesign.md, reverse-engineers any design system from a live site using npxskillui, and ensures every UI decision is grounded in explicit design intent rather than AI defaults.
- ▌ Kill Switch Binding · cleanexpoTelegram /panic command + dashboard kill button. Halts every role (CoS, Margot bridge, Scribe sends, Dispatcher mid-flow, Curator) within one cycle. Two-of-N + 2FA enforcement on dashboard side; Telegram side trusts only the configured operator chat_id. Closes the kill-switch safety control.
- ▌ Maintenance Manager · cleanexpoSenior Maintenance Manager (15+ years production systems). Evaluates dependency freshness, technical debt severity, upgrade paths, observability coverage, and produces a maintenance calendar with prioritised debt items and SLA recommendations.
- ▌ Readiness Architect · cleanexpoGenerates a complete SPM-ready readiness scope for any project, product, app, authority site, CRM, contractor network, training platform, marketing engine, mobile app, internal operating system, AI agent system, or launch-readiness task. Use when preparing a project for /spm, /judge, /goal, shipit, launch, production readiness, operational readiness, connector readiness, security readiness, or client purchase readiness.
- ▌ Remotion Production · cleanexpoUse when a Remotion video job needs production gates, render evidence, output discipline, and pass/fail validation before it can be called shippable.
- ▌ Resume From Handoff · cleanexpoResume work from a session handoff (/resume-from-handoff). Reads the latest handoff, verifies current repo state against it (read-only), reconciles drift, then continues the work from the documented pickup point without re-deriving old context. Verification is mandatory before any work resumes.
- ▌ Agentskills Manifest · cleanexpoExport Pi-CEO's skill registry as an agentskills.io-format manifest. Closes Hermes RA-1838 SWARM-009 (open question
- ▌ Launch Project Audit · cleanexpoScan the codebase and produce a plain-English map of every feature — what's built, stubbed, orphaned (built-disconnected), wired, or tested — plus how far each piece has travelled toward production. Cheap model for the bulk pass, frontier only for ambiguous calls. Use on "what's actually built", "scan the project", "what's left", or as step 2 of /ship-it.
- ▌ Marketing Copywriter · cleanexpoWrites long-form marketing copy — landing pages, blog posts, email sequences, ad copy, sales-page sections — strictly aligned to BrandConfig voice, positioning hierarchy, and ICP vocabulary. Use when a brief asks for "landing page", "blog post", "email sequence", "ad copy", "sales copy", "long-form copy". Reads positioning + ICP + channel-plan; never writes blind. Composes with marketing-seo-researcher (keywords) and remotion-orchestrator (CTA video assets).
- ▌ UI Component Builder · cleanexpoSenior-level UI implementation skill. Generates multi-variant React/Tailwind components grounded in DESIGN.md, applies taste-skill quality constraints (banning AI-default slop), and uses 21st.dev Magic patterns for component discovery and generation.
- ▌ Marketing Positioning · cleanexpoDevelops value proposition, competitive positioning, and Jobs-to-be-Done articulation for a brand or feature. Use when a brief asks for "positioning", "value prop", "messaging", "JTBD", "competitive analysis", "category creation", or before any landing page or campaign starts. Reads the brand's existing voice and audience from BrandConfig; produces positioning artifacts every downstream copywriter / strategist skill consumes.
- ▌ Pi Dev Ops Model Farm · cleanexpo bundleMulti-CLI agent farm orchestrator. Manages Claude Code (3x Max Plan accounts) and OpenAI Codex CLI (1x Max/Plus account) as tmux-based background workers. Receives routing requests from Pi-Dev-Ops /nexus, dispatches tasks, monitors health, and returns structured outputs via JSON-over-file IPC. Enables hybrid Claude+OpenAI execution without API billing.
- ▌ Pi SEO Health Monitor · cleanexpoPortfolio health trend analyst for Pi-SEO. Reads scan history from .harness/scan-results/, detects regressions and sustained degradation, identifies cross-repo systemic issues, computes portfolio health score, and generates alert digests.
- ▌
- ▌ Remotion Integrations · cleanexpoUse when /remotion-video needs existing Synthex ElevenLabs, brand-config, Remotion, output storage, Claude, or Hermes integration without adding vendors.
- ▌
- ▌ Terminal Orchestrator · cleanexpo bundleSafely manage long-running tmux-based dev workflows (build, test, watch, log tailing) on the local machine via a constrained command grammar. Use when an agent or operator wants to inspect tmux state, start a known profile, or run an allowlisted command inside a tracked session. NEVER use for production deploys, force-pushes, secret-exfiltrating commands, or anything matching the denylist. Read policy/ before executing.
- ▌ Launch Enhance Debloat · cleanexpoMake existing code stronger, leaner, and more secure without over-engineering — deletion is often the best change. Finds dead weight, weak spots on critical paths, and security issues; proposes a ranked reversible change list and applies only approved changes in a sandbox with tests passing. Use on "clean up", "remove bloat", "strengthen", "secure it", or as step 4 of /ship-it.
- ▌ Marketing Icp Research · cleanexpoBuilds an Ideal Customer Profile (ICP) — firmographics, role, pains, triggers, vocabulary, watering holes, decision process — for a brand or feature. Use when a brief asks for "ICP", "target audience", "buyer persona", "customer research", "user research", or before any campaign / copy / channel work. Reads BrandConfig.audience as the seed; outputs structured ICP consumed by copywriter, channel-strategist, seo-researcher.
- ▌ Marketing Orchestrator · cleanexpo bundleENTRY POINT for the Marketing Skills Package. Use the moment a brief mentions "marketing skills package", "marketing package", "use marketing", "campaign", "launch", "positioning", "go-to-market", "GTM", "marketing strategy", "marketing plan", or asks to produce marketing assets (landing page, ad creative, blog post, email sequence, launch runbook, social content) for one of the portfolio brands (DR, NRPG, RestoreAssist/RA, CARSI, CCW, Synthex, Unite) or any customer brand. Globally available at ~/.claude/skills/marketing-* (symlinked to ~/Pi-Dev-Ops/skills/marketing-*). Reads the brief, classifies (campaign type, brand, audience, channel mix, budget, timeline), emits a wave-plan JSON, dispatches sub-skills in topological order. Composes with the Remotion Skills Package — calls remotion-orchestrator for any video output.
- ▌ Pi Dev Linear Contract · cleanexpoBinding Pi-Dev-Ops × Linear contract — Pattern B repo→project routing, autonomy pickup requires status Ready for Pi-Dev plus pi-dev:autonomous label, run_id idempotency on every autonomous run, Pi-Dev workflow statuses, ticket footer format. Use before any Linear read/write by Pi-CEO code paths, autonomy poller work, or launch-crew sync. Triggers on "linear contract", "pi-dev linear", "Ready for Pi-Dev", "pi-dev:autonomous", "run_id", "Pattern B routing".
- ▌ Remotion Colour Family · cleanexpoGenerates a complete, accessible palette (primary, secondary, accent, 4-step neutral, semantic, and dark variant) from one to three hex anchors plus a colour-family classification. Validates WCAG-AA contrast for all text-on-background pairs. Triggered when a BrandConfig has fewer than 5 defined colours or contrast fails.
- ▌ Carsi Course Production · cleanexpoUse whenever creating, editing, refreshing, or reviewing ANY CARSI course, module, lesson, quiz, course trailer, or course marketing copy. Enforces the non-negotiable Australian-production standard (Australian English, 230V/10A power, metric units, Australian-available products, Australian standards/regulators, AUD) and the course quality bar.
- ▌ Design Family Coherence · cleanexpoCross-brand audit. Given a colour family (`restoration | safety | industrial | consumer | training`), reads every brand's `.design.md` + `.motion.md` and reports on family-level consistency — shared signature motion, shared spacing rhythm, shared typography weight ladder, palette saturation curve. Flags drift. Used after `design-approve` adds a new brand to the family, and as a standing monthly audit.
- ▌ Remotion Brand Research · cleanexpoResearches a portfolio company's visual identity from public sources (homepage, about page, marketing copy, GitHub README) and produces a structured BrandResearch dossier. Triggered when `Synthex/packages/brand-config/src/brands/{slug}.ts` (canonical home per RA-1985 / Synthex SYN-897) is missing, older than 90 days, or the user explicitly says "refresh brand for {brand}". Output feeds remotion-brand-codify.
- ▌ Unite Group CI Recovery · cleanexpoHow to ship PRs cleanly across the Unite-Group portfolio (Synthex, Pi-Dev-Ops, Disaster-Recovery, DR-NRPG, RestoreAssist, ATO, CARSI, CCW-CRM, Unite-Hub, synthex-mcp-app) without falling into the same CI / Vercel / Supabase / convention traps every session. Captures the recurring failure modes I hit 2026-05-25 across PRs
- ▌ Weekly Enhancement Loop · cleanexpoWeekly cross-repo self-improvement loop. Every Monday 02:00 AEST it applies the 8-Claude-Loops method (INGEST / BUILD / COMPOUND + North Star) across every repo in config/harness/projects.json, opening review PRs so all projects compound over time. Cost-optimised open-weight ladder (Qwen/DeepSeek) via OpenRouter — cheap grunt tiers, one reserved reasoner for the planner handoff (~$5/full run).
- ▌ Fleet Knowledge Pipeline · cleanexpoUse when asked to turn watched/read/recorded material into shipped work — "get this into the wiki", "research this and build it", "why isn't the fleet picking this up", "make the machines work on this" — or when a request spans learning → research → board decision → spec → build across the three-machine fleet. Also use before adding a new ingestion source, a new researcher, or a new board→build hop, so the existing chain is reused instead of duplicated.
- ▌ Margot Sandcastle Bridge · cleanexpoAfter a Margot Deep Research finding completes, classify "action-shaped vs informational" via existing intent_router, draft a Linear ticket body, route through draft_review HITL gate. On 👍, create the Linear issue with `sandcastle:high-isolation` + `pi-dev:autonomous` labels. The existing autonomy.py poller picks it up and the sandcastle-runner skill executes the work in an isolated container. Closes Wave 5
- ▌ Marketing Launch Runbook · cleanexpoBuilds a T-30 to T+30 product-launch runbook with a per-day timeline, owners, asset dependencies, gate checks, contingencies, and post-launch attribution. Use when a brief mentions "launch", "GTM", "go-to-market", "ship a product", "release campaign", "launch playbook". Reads campaign-plan + positioning + channel-plan + all content artifacts; calls remotion-orchestrator for launch-day video assets. Outputs the single source of truth that orchestrates every other skill on the calendar.
- ▌ Marketing SEO Researcher · cleanexpoPerforms keyword research, search-intent classification, SERP analysis, and content-gap discovery for a brand or topic cluster. Use when a brief asks for "SEO", "keywords", "search intent", "SERP", "content gap", "ranking strategy", or upstream of any blog / landing-page work that targets organic search. Outputs a structured SEO brief consumed by copywriter and channel-strategist.
- ▌ Marketing Social Content · cleanexpoWrites short-form social content — LinkedIn posts, X/Twitter threads, Instagram captions, TikTok hooks — and dispatches video shorts to remotion-orchestrator. Use when a brief asks for "social post", "LinkedIn post", "thread", "tweet", "Reels caption", "social campaign", "social calendar". Reads positioning + ICP vocabulary + channel-plan; cross-pack composes with remotion-orchestrator for any video output.