my-claude-skills
Health Uyari
- No license — Repository has no license file
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Gecti
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
- Permissions — No dangerous permissions requested
Bu listing icin henuz AI raporu yok.
Nine Claude Code skills I wrote and twelve third-party ones I use: spec, product check, research, design, review, project audit.
my-claude-skills
Skills for Claude Code — the ones I wrote, and the third-party ones
I actually use, kept in one place so my setup is reproducible.
Install any of them: copy the folder into ~/.claude/skills/<skill-name>/ and restart Claude Code.
Two folders, and the distinction is not cosmetic:
mine/— written by me, maintained here. See mine/README.md.external/— not mine. Third-party MIT skills, vendored with full attribution
and the original licence texts. Prefer installing those from upstream; see
external/README.md.
Mine
| Skill | What it does | Reach for it when |
|---|---|---|
| idea-storm | Nine options across three shelves, three finalists, then two blind reviewers and a decision with a kill sign | A choice is open and the first sensible answer is not good enough |
| spec-first | Project kickoff in two passes: Vision (what and why), then Blueprint (how and in what order) | A new project exists as an idea and nothing is written down yet |
| product-check | Five jobs-to-be-done questions, then a readiness score with risks | The idea exists and the real question is whether to build it at all |
| deep-research | Multi-agent research with web search, confidence ratings, saved artifact | The answer is worth several sources and being wrong is expensive |
| design-first | Reads the look from the topic, six first-screen mockups drawn by an image model, you pick one, then a live page is built over it — desktop and phone each from their own mockup, checked with numbers | Visual work is starting and "make it look good" is not a brief |
| multi-layer-review | Up to 5 blind parallel reviewers: architecture, code, user POV, robustness, requirements | A spec is finished and you want it attacked before code exists |
| project-audit | Audits CLAUDE.md quality, memory consistency, file structure, git hygiene | A project has grown messy, or before handing it to someone |
| human-text | Russian copy that reads as human-written — rhythm, honest hedging, no AI clichés | Anything a person will read: posts, articles, landing copy, letters |
| reels-console | A vertical reel cut from one static take: pauses gone, crop levels anchored to the face, word-timed captions, music ducked | You filmed yourself talking to camera and need it edited without opening CapCut |
🇷🇺 У каждого скилла внутри есть описание на русском.
External — not my work
Vendored third-party skills, all MIT, each with author and source in
external/README.md and licence texts in
external/licenses/.
| Skill | What it does | Author |
|---|---|---|
| pricing | Value metric, tiers, willingness-to-pay research, when and how to raise prices | Corey Haines |
| marketing-plan | 12-month plan across the funnel, 13 sections, 139-tactic library included | Corey Haines |
| site-architecture | Page hierarchy, navigation, URL structure, redirects when restructuring | Corey Haines |
| negotiation | Chris Voss method: mirroring, labelling, calibrated questions, Ackerman ladder | Wondel.ai |
| high-output-management | Grove: leverage, limiting step, indicators that cannot be gamed, 1:1s, OKRs | Wondel.ai |
| doc-to-markdown | DOCX/PDF/PPTX to clean Markdown, tables and images intact | daymade † |
| mermaid-tools | Mermaid diagrams out of Markdown into high-resolution PNG | daymade † |
| ppt-creator | Decks with assertion-style headings, charts, speaker notes, self-scoring | daymade |
| meeting-minutes-taker | Transcript to minutes without losing content; honest about who spoke | daymade |
| youtube-downloader | yt-dlp plus the bot-check, 403 and locked-platform handling | daymade |
| ui-ux-pro-max | Design reference as queryable data: styles, palettes, type, UX rules, 13 stacks | nextlevelbuilder |
| skill-conductor | Skill lifecycle: create, improve, validate, review, optimise, package | smixs |
† Patched, not byte-identical to upstream — both were broken on Windows. Each patch and the reason
for it is documented at the top of that skill's README.
Not vendored on purpose: remotion-best-practices (Remotion team) is excellent, but
remotion-dev/skills ships no licence file at all — default
copyright, all rights reserved. Redistributing it would not be legal. Install it from the source.
Why they are built this way
Two of the skills ship the research they came out of, as research.md inside the skill folder:
- mine/spec-first/research.md — 10 methodologies, 12 templates and 6
real project disasters, boiled down into the VIBE SPEC approach. The argument for why a spec is
worth writing when you are not writing the code yourself. - mine/design-first/research.md — the June 2026 research behind the
vibe reading and the numeric check (archived). The switch to image-model mockups and the blind
rounds behind it are summarised in references/mockups.md.
🇷🇺 Обоснования лежат рядом со своими скиллами, а не отдельной папкой — «почему так сделано»
логичнее искать там же, где сам скилл.
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi