Shape agent capabilities into your workflow.

cinematic-ai-film teaches an end-to-end workflow for finishing an AI promo film, from locked beats and image-to-video shots to one consistent voice, real product UI capture, sound design, ffmpeg assembly, and measured overlays. It focuses on the last-mile failures that ruin a deliverable—continuity drift, mispronunciations, mask alignment, and runaway renders—and records costs and hard-won fixes from a real launch film.

elevenlabs-v3-text-optimization converts flat text into ElevenLabs V3-ready scripts with emotional audio tags, pacing, punctuation, and contextual formatting. It respells names and brands for pronunciation, expands acronyms and numbers, removes unsupported SSML, and adapts the approach for dialogue, narration, technical, or dramatic delivery. Use it when the same words need clearer timing and more reliable synthesized speech.

Build a reusable candidate knowledge base from your resume, portfolio, projects, and past interviews, then research the company, role, and product against the job description. The skill identifies weak or unclear evidence, drafts likely HR and hiring-manager questions with answers grounded in your real experience, and prepares a natural outreach message. Fresh company or interview information still depends on web access or materials you provide.

de-ai-gate scans HTML, prose, and authored records for typographic and phrase patterns associated with machine-written copy, then exits clean or failed with a deterministic code. It writes nothing, uses no network or credentials, and can run from Node, Bun, or Python in a pre-ship or CI gate. It reports offending tells without modifying the files.

Slopspeare is a field guide for spotting and replacing patterns that make prose feel machine-written: inflated vocabulary, contrast pivots, tidy three-part lists, excessive em dashes, and over-neat conclusions. It explains why each rule matters and provides a printable final-pass checklist plus quick reference. The guide applies to bios, landing pages, case studies, docs, blogs, social posts, and release notes.

motion-ui checks a project’s framework, package manager, Tailwind setup, and existing motion dependency before changing anything. It installs the appropriate motion package and builds 21st.dev-style components through Magic MCP or local recipes. Confirmation mode is the default; an opt-in auto mode runs end to end, and unsupported stacks receive an alternative instead of a forced install.

localize-infographic-pdf uses a hidden searchable text layer to locate visible vector or Type3 text, erases each source run, and redraws the translation in the same position, color, weight, and size. Photos, charts, gradients, icons, and logos are not regenerated. It targets Canva, Figma, or Illustrator one-pagers and follows analyze, extract, compose, and PDF stages with translator and reviewer agents.

Use this methodology to turn an AI initiative into a measurable, stage-gated plan that can stop before expensive product work begins. It decomposes capabilities, ranks risks, isolates one major variable per stage, defines human validation and acceptance metrics, and sets explicit pass, fail, and stop-loss conditions. The central rule is to prove the core AI capability before building the surrounding platform, interface, database, or automation.

Give reeljet real screen recordings, screenshots, a logo, and a brief to produce a directed short-form product ad. It creates the hook, storyboard, music choice, and estimated generation cost before an approval gate, then combines Higgsfield-generated motion with the real footage using captions and ffmpeg. The final campaign includes both 16:9 and reframed 9:16 videos and requires Claude Code, Higgsfield MCP access, Python, and ffmpeg.

share-html is a collection of standalone HTML templates for reports and interactive briefs. The editorial style includes a sticky table of contents, searchable tables, Kanban boards, charts, and progress checklists, while the handdrawn style uses Traditional Chinese headings, Rough.js borders, notes, and whiteboard blocks. Because data is embedded, rendered files can open from file:// or on mobile without a server.

bug-scanner reviews the current Git diff by default or scans paths you name, looking for null access, missing awaits, races, injection, missing authorization, SSRF, unsafe deserialization, N+1 queries, blocking I/O, and similar defects. Findings are grouped from Critical to Low with file locations and concrete fixes. It only reports issues with a real consequence, not naming, quote-style, or formatting nits.

Open an HTML file through YapUI, then describe changes by typing, speaking, selecting an element, taking a screenshot, or recording the screen. A local Node relay sends each note to a resident Claude agent, streams its status and reply into the page, and refreshes the result after edits. Instant mode requires the Claude Code CLI; without it, the same interface falls back to a watcher driven by the main session.

Give Riffkit a successful TikTok or uploaded clip and it will study the hook, pacing, and emotional beats before generating new footage around your product, character, and story. The hosted service can return voiceover, timed captions, a cover, post copy, and hashtags in English or Spanish without re-uploading the source footage. Generation requires a Riffkit account and is billed by the finished video's duration.

the-super-planner-skill sizes the task as Quick, Standard, or Deep, routes research by claim type, and checks the project’s dependency versions before relying on current documentation. It locks a Goal Contract with numbered success criteria and a runnable definition of done, then traces each implementation slice back to one criterion and repairs failed checks. The skill has no extra dependencies, but Claude Code details can change with installed versions.

Install the skill in Claude Code to get focused guidance for Unity 6 URP 2D work, including sprites, tilemaps, sorting, batching, physics, lighting, and pixel-art setup. Its reference files trace rules to official Unity documentation and separately flag behavior confirmed in the editor when defaults or common assumptions differ. It is most useful for concrete implementation and performance decisions within Unity 2D.

verified-experiments adds a Python harness and Claude skill with guards for data and Git provenance, real-data sanity, label leakage, impossible metrics, stored predictions, reproducibility, silent fallbacks, guard meta-tests, and code review. Each guard rejects a deliberate fake, and the diagnoser explains the failed gate and next step. An optional agent loop may edit experiment code but cannot change the guards, tests, reviewer, or diagnoser to make a run green.

resume-radar simulates an LLM resume screener with a reverse-engineered 120-point rubric, including a GitHub open-source classification that can affect the score. It returns a gap table sorted by point swing and concrete edits, while refusing to fabricate experience, pull requests, stars, or metrics. Use it to understand machine-screening signals, not to inflate a candidate’s record.

Claude Code Doctor inspects the workspace layer that shapes every session, including context files, permissions, skills, MCP tools, automations, and disk usage. It produces a graded Markdown report and local HTML dashboard with red flags and concrete prescriptions. The diagnostic pass is read-only apart from the approved report destination, and cleanup actions are separated behind explicit user approval.