makerskills
AI agent skills for the personal operator's craft. Decisions, research, second-brain workflows, content rotation, scenario modeling, CFO cadence, domain hunts, and the meta-skills to author more.
Built for founders and indie operators. Works with Claude Code, Codex, Cursor, and other Agent Skills hosts.
Site: maker-skills.com
Install
# In Claude Code
/plugin marketplace add coreyhaines31/makerskills
/plugin install makerskills@makerskills
Or symlink for local dev:
git clone https://github.com/coreyhaines31/makerskills ~/code/makerskills
ln -s ~/code/makerskills ~/.claude/plugins/makerskills
See INSTALL.md for env vars, personal-config setup, and runtime dependencies.
Getting started — 5-minute path
New to the plugin? Do these in order. Everything else can wait until you need it.
1. Install the plugin (above), then set the base env var in ~/.zshenv:
export MAKERSKILLS_CONFIG="$HOME/.config/makerskills"
Reload your shell.
2. Run your first skill — start with decide. No config required:
/decide
It'll ask what decision you're facing, pick the right questions from the 37signals framework, and archive the result with a revisit date. That's the first-touch experience: any skill in this plugin. Structured input → structured output → written to disk.
3. Read one SKILL.md to see the pattern:
open ~/code/makerskills/skills/decide/SKILL.md
Each skill is a workflow doc — you can invoke via /decide OR read the SKILL.md and run the steps by hand. The plugin is documentation-first; the automation is a side effect.
4. Try one more skill that involves personal config, so you understand the pattern:
/paste twitter
It'll read your clipboard, strip formatting, warn if over 280 chars, and copy back the cleaned output. First invocation may prompt you to install pbcopy-adjacent deps if missing. No config file needed — this is a pure utility.
5. Now branch out. Skim The 22 skills below and pick one that maps to a workflow you're already doing manually. That's the highest-leverage adoption path.
Which skill do I invoke when?
Routing table for common operator jobs. Match your intent → skill.
| I want to... | Skill |
|---|
| Think through a decision with a real fork | decide |
| Break through a wall ("this seems impossible") | unstuck |
| Get a board of famous founders to weigh in | maker-council |
| Pressure-test a new business or product idea | business-brainstorm |
| Research a topic with citations | deep-research |
Find an available .com for a new project | domain |
| Watch known sources daily and feed the good stuff into my brain | radar |
| Capture / query / lint my personal knowledge base | second-brain |
| Same but for a team-shared knowledge base | company-brain |
| Extract notes / highlights / summaries from a book | |
Not sure between two? The skill's SKILL.md description always includes trigger phrases and disambiguation from adjacent skills.
The 22 skills
Meta — extend Claude Code (the -ify trifecta)
| Skill | What |
|---|
skillify | Create, adapt, or update a skill in any sibling repo. Three modes: CREATE (from-chat / from-video / from-dump / from-scratch), ADAPT (port external skill with license check + attribution), UPDATE (improve existing skills with cross-skill propagation + semver discipline), TEST (role-play users through a skill and judge transcripts against a rubric). |
toolify | Wire up an integration, API, MCP server, or third-party service into a Next.js or Rails project. Interactive wizard for auth, env vars, client wrapper, webhook handling, and smoke-test. |
loopify | Set up an agent loop, cron-scheduled task, or recurring workflow. Judgment layer on top of ScheduleWakeup / CronCreate / /loop — picks pattern (dynamic / cron / one-shot), tunes delay for cache windows, enforces idempotency + bail-out conditions. |
Decision & strategy
| Skill | What |
|---|
decide | 37signals decision framework (38 questions + house additions like Q39 opportunity cost, triaged to 6–8 per decision). Archives with revisit dates. |
business-brainstorm | Pressure-test a new business or product on 9 dimensions. Composes with deep-research + domain. |
domain | 11-step .com domain hunt on Laura Roeder's "availability-first, never fall in love with a name" methodology. Multi-tool ensemble: Vercel CLI + whois (per-TLD) + Domainr API + Namecheap API + rdap.org (modern TLDs) + agent-browser for USPTO trademark screening + aftermarket click-throughs (HugeDomains / Afternic / Sedo / Dan). Handles bare-word + prefix/suffix brainstorming, budget filtering, negotiation guidance, and social handle checks. |
deep-research | Multi-source research with citations + archive. WebSearch, WebFetch, agent-browser, /last30days, memory. |
unstuck | The roadblock antidote. Classifies what kind of "no" you hit (assumption / framing / gatekeeper / tool / resource / physics), runs targeted lateral-thinking techniques from a 10-technique inventory, 10-angle minimum before evaluating. Agents run it on themselves before reporting any dead end. Upstream of decide. |
maker-council | Simulated personal board of advisors — Fried, Musk, Bezos, Jensen, Iger, Graham, Naval, Blakely, Cohen. Seats 3–5 by question type with a designated dissenter, grounds takes in documented frameworks (optional live-research pass), maps the disagreements, synthesizes a call. Custom members via config dir. |
Knowledge & content consumption
| Skill | What |
|---|
second-brain | Karpathy LLM Wiki workflow over any markdown vault. Capture / compile / query / lint / connect / search. Personal-scope. |
company-brain | Team-scope sibling to second-brain. Structured raw dirs (people / companies / meetings / sops / decisions / customer-language / recurring-questions / sales-objections), multi-author attribution, sensitivity tagging, trust levels + a /cb review culling pass so unreviewed or deprecated info never poisons answers, optional auto-sync from Fathom / Gong / Granola / CRM. Backbone for a Company Brain Setup productized service. |
radar | Standing surveillance on known sources — YouTube channels, RSS/blogs, subreddits, Hacker News, X, LinkedIn, keyword searches. Polls on a daily schedule (launchd or systemd), fetches only what's new (per-source state files), scores it against your stated focus, writes one digest to the vault, and auto-captures only what clears the bar. Everything else waits in the digest until you promote it. Fills raw/ for second-brain to compile. |
read-book | PDFs, EPUBs, MOBI, markdown — chapter-by-chapter notes, quotes, summaries, or spaced-rep study mode. |
watch-video | YouTube, Loom, Vimeo, Riverside, Zoom, MP4. Transcript / visual / multimodal (Gemini-native) modes. |
ingest | Raw human input — call transcripts (Grain/Zoom/Granola/Fathom), client texts, emails, voice notes — extracted into decisions, action items (yours vs theirs), filed GitHub issues, vault captures, and a drafted-never-sent reply. Person→project routing via private . |
Output & creative
| Skill | What |
|---|
jab-hook | Gary Vee's jab-jab-jab-right-hook rhythm applied to your portfolio rotation on X + LinkedIn via Typefully. |
slide-deck | Branded React decks (Next.js). "Show, don't tell" narrative pitching, density modes, PPT conversion, Playwright export. |
Operations & utilities
| Skill | What |
|---|
pm | Kanban (5-column) + Eisenhower across your portfolio. Tool-agnostic adapters (Notion, GitHub, Plane, Linear, Obsidian, manual). |
personal-cfo | Personal financial scenarios — house purchase + rental forecasting, monthly cash flow, big-purchase decisions. Household scope. |
company-cfo | Company / agency CFO workflow — monthly snapshots, weekly cash pulse, scenario projector. Transaction-sum EOM methodology, categorization discipline, forward forecasting. Pulls from bank + payment processor + payroll + expense-mgmt via toolify-wired integrations. Team scope. |
paste | Clean terminal output for 9 destinations (Slack, Notion, Twitter, LinkedIn, email, GitHub, plain, HTML). Secret-detection first. |
social-fetch | Pull any social post by URL across 10 platforms (X, LinkedIn, IG, TikTok, Bluesky, Reddit, Mastodon, Threads, HN). Strategy chain with free + paid fallbacks. |
Skills compose
Skills call each other by name. When a skill's job hits an adjacent job, it routes there rather than reimplementing. The most-referenced skills are the "central" ones — expect to touch these first as you adopt:
| Most referenced by siblings | Composes with... |
|---|
watch-video (54 refs) | second-brain (capture summary), skillify from-video, social-fetch (X/IG/TikTok metadata), pm (action items), decide (flagged decisions) |
skillify (51 refs) | watch-video (record → skill), second-brain (capture source), compound-engineering:* (Anthropic-official skill authoring) |
second-brain (48 refs) | deep-research (external gaps), read-book (highlights → wiki), watch-video (summaries → raw), paste (clean captures) |
slide-deck (38 refs) | business-brainstorm (pitch decks), watch-video (talk-recording → outline), second-brain (wiki as content) |
company-brain (33 refs) | toolify (wire auto-sync), loopify (schedule sync), deep-research (external gaps), decide (structured archive) |
domain (32 refs) | business-brainstorm (name after idea passes filter), toolify (wire Domainr + Namecheap), decide (candidate fork) |
See each skill's ## Composes with section for the full dependency list.
Architecture
The repo is public + generic. Your personal data (Typefully workspace IDs, portfolio properties, voice overlays, archives) lives separately in ~/.config/makerskills/ (gitignored, on disk only).
Skills read from ${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/<skill>/ paths. Set env vars in ~/.zshenv to point at your locations:
export MAKERSKILLS_CONFIG="$HOME/.config/makerskills"
export SECOND_BRAIN_VAULT="$HOME/Documents/SecondBrain" # for second-brain (personal-scope vault)
export COMPANY_BRAIN_VAULT="$HOME/Documents/CompanyBrain" # for company-brain (team-scope vault)
export COMPANY_CFO_ROOT="$HOME/code/company-cfo" # for company-cfo (reports, projector, config)
export SLIDE_DECK_REPO="$HOME/code/your-site-repo" # for slide-deck
Full setup in INSTALL.md. Full mental model in ARCHITECTURE.md.
Versioning
Two levels of semver:
- Plugin release tag (e.g.
v0.5.0) — this repo's overall version. Bumped on each meaningful addition to the collection. See CHANGELOG.md.
- Per-skill
metadata.version in each SKILL.md frontmatter — that skill's individual version. Evolves independently.
A plugin release at v0.5.0 may contain skills at 0.1.0, 0.1.1, 0.2.0, 0.3.1 simultaneously. That's expected — each skill has its own iteration cadence.
SKILL.md format
---
name: skill-name
description: When the user wants to [X]. Triggers on "[trigger phrase]"...
metadata:
version: 0.1.0
---
The description is what Claude uses to decide when to load the skill — be specific about triggers and disambiguation from sibling skills.
Cross-referencing other skill packs
Skills can reference each other across plugins. In a SKILL.md description or body:
- Within this plugin: mention by name (
see decide)
- Cross-plugin: prefix with the plugin (
see marketingskills:cro)
Claude resolves references by description match at load time — no manifest linking needed.
Related skill packs
Docs
INSTALL.md — env vars, personal config setup, runtime dependenciesEXAMPLES.md — one worked example per skill: what you say → what the skill producesARCHITECTURE.md — mental model, families, personal-config pattern, sibling-repo ecosystemCHANGELOG.md — release history with per-version detailsCONTRIBUTING.md — how to add a skill, testing, PR conventionsFAQ.md — common questions and gotchasBACKLOG.md — planned skills + brainstorm candidates
License
MIT — see LICENSE.
Built by
Corey Haines — founder, Conversion Factory + Magister.
