/ AI systems and agents
Skill library.
66 Claude Code skills plus the playbooks they run on. These are the verbs the whole operating system speaks.
/ Client work · 21
Onboarding deckDuplicates and personalizes the client onboarding deck in one command.
Onboarding script batchDrafts a new client's first three scripts from their onboarding notes.
Post-call emailDrafts the recap email after every client call, matched to the client's stage.
Content-ready emailDrafts the delivery email the moment a client's batch is ready.
Welcome letter writerWrites the handwritten-style thank-you letter mailed to every new client.
Client performance reportMonthly performance report: verified numbers from the analytics API, sheet plus deck.
Bulk post drafterDrafts a month of posts across a client's platforms in one run.
Bio writerWrites a client's social bios per platform, inside exact character limits.
Titles and pinned commentsYouTube titles and pinned comments generated from client scripts.
Shot list builderTurns a script into a filming shot list, sorted by who films what.
Meeting agenda keeperKeeps a living agenda page per client: next call, parking lot, covered.
Content ideation engine50 grounded content ideas per run, crossed against what people actually ask online.
VSL writerWrites video sales letters in the founder's own voice.
Viral hook creatorWrites and grades hooks against a library of proven ones.
Account optimizationPer-platform account optimization checklist plus the client walkthrough deck.
Video breakdownFrame-by-frame teardown of any video: hooks, structure, why it works.
AI UGC factoryGenerates UGC-style talking-head video clips in 4K vertical.
Meta research promptsReady-to-paste prompts for Meta's AI: influencer research, react-to videos, brand mentions.
Caption writerSEO-aware video captions in the founder's voice, written from the script itself.
Weekly client emailOne weekly performance email per active client, drafted, never sent.
Posting runwayMonday flag for any client about to run out of scheduled posts.
Hourly drafterReads the inbox every hour, drafts replies, files tasks. Never sends.
Morning agentReads five sources every morning, drafts what's owed, DMs the day's top three.
Weekly founders emailDrafts the Wednesday status email: team todos, client status, AI news.
Quality graderSundays: grades every draft against what actually got sent, proposes edits to the drafting logic.
OS maintainerWednesdays: audits the whole system for drift and ships approved fixes.
FAQ learnerMondays: harvests genuinely new answers from a week of sent mail.
Forms harvestFiles monthly client form responses to the right client automatically.
Email response drafterThe voice-matched drafting logic every email agent shares. Drafts only.
Email triageThe inbox-triage logic that now powers the hourly drafter.
Work-side executorChief of staff over the work fleet: morning brief, task execution, bridge reports.
Bug captureLogs a skill bug the moment it happens.
Bug queueRenders the open bug queue.
Apply fixesWalks the bug queue and fixes, one approval at a time.
Verify fixConfirms each fix works, or reverts it.
Build captureLogs every new build to the builds database automatically, in pure Python without an LLM in the loop.
Inventory regeneratorRebuilds the skills-and-agents inventory from live sources, weekly.
Impact rendererRenders the impact dashboard as a polished, shareable page.
System map generatorDraws an interactive map of the entire automation system.
Session wrapWrites the handoff when a session ends, one per work stream.
Session pickupStarts the next session with full context from the last one.
Mid-session syncPushes work to GitHub mid-session, behind a review gate.
Todo purgeWeekly cleanup of completed tasks. Pure REST, zero LLM tokens.
SenseiDescribe the task, get pointed at the right skill.
No-bullshitEnforces the honesty floor, so claims get verified before they ship and uncertainty is labeled inline.
Plain EnglishRewrites any overlong reply into short plain English.
FitfoForces unconventional options when the obvious ones fail, then commits to one.
Cofounder modeStrategic sparring partner with a running decision log.
Pressure testGrades a business idea across five areas against the Startup Playbook. Not a yes-machine.
Build modeA teaching mode where every change comes with a plain-English note on the one design choice that mattered.
Working with ClaudeAnthropic's collaboration best practices, distilled and always on.
AI writing QAScrubs scripts and copy for AI tells before they ship, then rewrites them clean.
Design standardsAudits any UI against written interface guidelines.
Video standardsHolds programmatic video work in React to a written standard.
Reality checkThe verify gate: every done, sent, or broken claim re-checked against live state first.
LeanToken-discipline mode, proven 26% cheaper on big builds in a 24-run blind test.
Flow mapAudits every automation from live state and draws the data-flow map.
Publish skillScrubs a private skill public, behind a fail-closed leak scanner.
iMessageReads and drafts texts locally. Drafts only, always.
Swipe deckPoint it at a folder or unread texts, swipe each card, actions run in a reversible batch.
Birthday textsWeekly prompt listing upcoming birthdays; reply with what to say and it drafts each text.
Course pullerFiles a full online course into a personal study library, transcripts and all.
Personal orchestratorDaily digests, promise tracking, verified briefs, always on.
Owed ledgerTracks every promise made by text or email, and auto-closes the kept ones.
Sculpt class plannerBuilds a complete, music-timed yoga sculpt class for the month. Ships as a plugin.
Sculpt setupOne-time onboarding that tunes the class builder to a new instructor.
Text triageWhich incoming texts matter, which wait for the digest, and which stay untouched.
Finance intakeReceipt-to-ledger flow for the company account: categorize, flag deductions, monthly rollup.
The bridgeThe append-only protocol the two agents use to hand work across the wall.
Website patrolWeekly stale-content sweep of this very site, drafts fixes, ships on approval.
Notion patrolWatches the workspace for pileups and rot, proposes structure instead of imposing it.
Desktop patrolCounts the mess, proposes the cleanup, touches nothing without a yes.
Stack advisorTracks tool and model releases against the live stack, recommends with sources.
Email opsInbox rules for the personal side: what gets drafted, labeled, or left alone.
LinkedIn deskStrategy, voice system, performance loop, and a daily 8:45am pack: comment replies drafted, the day's post ready, numbers logged Mondays.
One entry for every skill running in the live system, each built to keep a person in the loop.