cgbarlow-skills

Drive the iris-cli command-line tool to search, browse, ask AI questions, and export content from an Iris architecture-knowledge repository (https://github.com/cgbarlow/iris). Triggers on iris/diagram/element/package/set/collection language and on `iris …` commands; walks install (uv tool), the SQLite-vs-Supabase login split, and the read/export/ask command surface.

File a New Zealand individual IR3 income tax return for the 2026 tax year by driving the Inland Revenue myIR portal in the browser. Pauses for RealMe sign-in, refuses to click final Submit without explicit fresh consent, grounded in the official IR3G (March 2026) guide.

Interactive timesheet entry — collects work description, date, and start/end times via prompts, calculates hours, and appends rows to `.campaign/timesheet.md`. Supports clock-in / clock-out shortcuts.

Convert a PDF file to ebook formats (epub, mobi, azw3/Kindle) using Calibre.

Interactive prompt-engineering coach that elevates vague prompts through Socratic dialogue, multiple transformation styles, and guided learning.

Interactive claude-flow orchestration coach that guides users through swarm topology selection, agent deployment, memory configuration, and SPARC workflows.

Coaching companion for a 13-year-old advanced math student. Triggers on math homework, screenshots of questions, or 'I'm stuck' moments. Refuses to give answers, hooks interest with basketball / anime / advanced-math connections, and gently catches him when he has skipped the workbook instructions.

Six psychologically-grounded team role agents (Bear, Cat, Owl, Puppy, Rabbit, Wolf) plus Simon as educator/supervisor. Prerequisite for the full Campaign Mode experience.

Quest-based extension for AI-assisted work. Three NPC agents (Gandalf, Dragon, Guardian) provide mentorship, adversarial testing, and quality gates for structured campaigns.

Answer outcomes-theory questions strictly from Dr Paul Duignan's DoView Planning and Outcomes Theory Handbook (2025). Faithful adaptation of Prompt A v1.1.9 from https://www.doviewplanning.org/bookai. Triggers on outcomes-theory questions, requests for a DoView analysis, asks to critique a strategy/proposal/plan/policy/programme through an outcomes lens, or mentions of outcomes theory/outcomes systems/DoView Boards/theory-of-change. Produces the prompt's required formal Summary + Full standalone structure with raw-visible URLs, no markdown links, the outcomes-system definition rule, and an image-retrieval seed list. Pairs with doview-image-retriever.

Retrieve and faithfully reproduce DoView Planning handbook diagrams to accompany an outcomes-theory answer. Faithful adaptation of Prompt B v1.1.9 from https://www.doviewplanning.org/bookai with a Mermaid-first overlay — retrieves Mermaid blocks from the doview-book Markdown edition (https://github.com/cgbarlow/doview-book) first, falling back to upstream PNG/image-file URLs for diagrams that don't translate cleanly to Mermaid. Triggers after a doview-outcomes-answer response when the user asks to see the diagrams, requests the visual model, or says 'show me the picture'. Never redraws, simplifies, or substitutes images.

Phased orchestrator (scripts/shop.sh) for the weekly Woolworths NZ online grocery shop. Three phases: (1) Claude OCRs a photo of your meal plan and writes it to Iris via the Iris MCP, (2) pure-bash bulk-add uses cached SKUs stored in the Product attribute notes of each Iris Ingredient element — zero LLM tokens, (3) the woolies-shopper skill resolves any exceptions left over (out-of-stock, no cached SKU, etc.) and writes resolved SKUs back to Iris for next time. Run `./scripts/shop.sh` from a terminal. The skill itself acts as the phase-3 exception resolver — for the full weekly shop use shop.sh, not the skill directly. Pairs with the Iris meal-plan + aggregate workflow; requires iris >= v6.31.0 (ADR-217 provenance flag) for the cache fast path, degrades gracefully without.

AI-led 6-step sprint that walks an AI builder through validating one idea — vertical, buyers, framework, positioning — to a sales deck, outreach plan, and pre-seed investor deck. Adversarial by default; output is artifacts, not product. v0.3.2 makes explicit that the NotebookLM bolt-on is Claude Code CLI only and does NOT work in Claude Cowork (Cowork users use Claude-only synthesis, which works fully). Depends on six-animals + campaign-mode.