Start with the job, then find an open-source answer

mikeross turns OCR’d PDFs or document collections into an interlinked Obsidian vault where every note cites its source page. Its question mode answers only from the vault or admits that the material is missing, while mapping and graph modes help organize large collections and flag gaps. The skill is designed for grounded research with pypdf required and optional OCR support for scans.

visual-screen-flow lays real, runnable components from a repository on an infinite canvas so multi-step and branching UI flows can be inspected together. A separate flow JSON stores arrows, notes, lanes, and positions, while stable references resolve each node to the actual component without embedding source code. It is framework-agnostic and suited to onboarding, checkout, auth, and other state-heavy prototypes.

claude-conversion-copy diagnoses an audience’s awareness, desires, beliefs, and exposure to competing pitches before choosing how to write. In write mode it produces sales pages, landing pages, ads, emails, VSL scripts, headlines, and CTAs; in diagnose mode it finds the biggest leak, supplies a rewrite, and ranks fixes. The skill is built for conversion work that needs judgment, not a generic headline formula.

EXPOUNDER turns a napkin-thin project brief into a layered specification without drifting into a different idea. It first asks permission, then uses three to five sharp questions to anchor the north star, scope, platform, and key distinction. The spec records assumptions, stops only for decisions that would change the project, and ends with a build handoff containing a kickoff prompt and first tasks.

png-to-drawio reconstructs a PNG or AI-generated architecture diagram as editable draw.io vector shapes and can also emit a native PowerPoint with real objects for boxes, icons, and labels. It favors recreation over importing a flat image, preserving structure for versioning, editing, and re-export. The skill is useful when a polished raster diagram needs to become an engineering artifact.

spotlight-intro animates a full list as a waterfall of text, pauses one chosen item at center, then zooms it through the frame into a branded solid-color end card. It renders a real MP4 on a single canvas with WebCodecs and mp4-muxer, supporting landscape and portrait formats. The setup needs the complete list, featured item, two brand colors, background, and what follows the opener.

geo-content-forge audits a live page, classifies keyword opportunities from search signals, and writes paste-ready Title, Meta, H1, question-led H2s, FAQs, and FAQPage JSON-LD. It compares each element with the live page, labels facts by their source status, and blocks guesses, invented search volume, and ranking promises. A human-voice check also screens the output for common AI writing tells.

idea-discovery-framework guides a product idea through nine structured sections, scores commercial viability and solo-buildability against a consistent rubric, and returns a kill-or-proceed decision with explicit reasoning. It reads an idea brief, project instructions, and a builder capability profile, then writes each evaluation into a separate workspace.

moneyfrisk checks changed code for floating-point currency arithmetic, late rounding, and amounts written directly to a ledger. It points to the file and line, suggests integer cents or an explicit rounding policy, and keeps ambiguous tax decisions open for you to choose. Use it on checkout, billing, tax, balance, or settlement changes where a tiny numeric error becomes a real discrepancy.

research-reports combines a report-writing discipline with a research cache and a policy-source guide. It designs scope and terminology, stores full source text alongside URLs and local paths, grades evidence, declares boundaries, and reviews multiple sources before finalizing. The policy skill routes US sanctions, regulatory, and policy work to allowlisted domains and official APIs. It is for professional multi-source reports, not casual fact lookups.

transcript-fix works after speech-to-text: it compares two to four transcripts of the same medical talk, reconstructs likely names, drugs, numbers, and dosages, and returns a cleaned transcript plus unresolved items. It never fills an uncertain detail just to make the prose smooth. Only corrections you personally confirm or verify against the literature enter the growing Notion glossary, so shared transcription mistakes do not become accepted terms.

tighten removes filler connectives, throat-clearing, hedging, empty intensifiers, repeated context, and other padding while preserving every fact. Numbers, paths, code blocks, error strings, decisions, and caveats stay intact. It can edit a file or section in place or return a rewrite for pasted text; the goal is tighter prose, not summarization or a different argument.

meta-analysis-skill coordinates a protocol, PRISMA-S search ingestion, dual-blind screening, dual extraction, local R statistics, forest and funnel plots, and an APA or PRISMA-2020 manuscript draft. Seven sign-off gates stop at decisions that carry authorship or interpretation. The bundle also checks citations and typesets the manuscript. Its demo uses synthetic data, and the repository is source-available under non-commercial licenses.

animeblip-showrunner-skill drives Animeblip from a story brief through screenplay, characters and recurring elements, scene images, scene videos, and a final cut. It can run end to end or pause between stages for review and targeted fixes, and it supports shorts, films, and episodic series. The skill itself uses the fixed Anime6 style; other styles must be changed in the web editor. An Animeblip API key and hosted service are required.

osint-field-graph defines a field and disqualifier, seeds an initial set from GitHub topics, OpenAlex, or a supplied edge file, then expands the graph with forward BFS and ranks nodes across sources. Persona cross-checking and a domain filter narrow the result. The core is pure shell, and GitHub mode needs authenticated gh. Its scope is limited to public accounts, public relationship graphs, and signed work; it is not a deanonymization tool.

cockpit aggregates the sources you wire in—CLIs, MCP servers, or local files—then ranks items by urgency, convergence across surfaces, and stakes. It renders a single HTML brief with the data embedded, so the file opens from file:// or on mobile without a server. The skill is intentionally read-only: it never sends, moves, or writes anything, and the standalone renderer needs only Python 3.

naturalize uses a SessionStart hook to nudge Claude toward natural prose throughout a session while skipping code, commits, and file paths. Each run rewrites the draft, rereads it as a skeptic, fixes passages that still sound synthetic, and ends by removing em dashes and en dashes. You can install it as an on-demand skill instead, and its test corpus covers the cataloged patterns.

audit-croatian-gdpr-compliance-skill classifies the product, stage, data subjects, data categories, and controller or processor role before applying the EU GDPR baseline and Croatia-specific gates. Each check is marked pass, fail, risk, not applicable, or needs legal confirmation, with observed basis and article references in a self-contained HTML report. Its source register is dated, so refresh official sources before relying on high-stakes or client-facing conclusions.