Joshua Nowak
Community@JYeswak · Missoula, Montana
Spent 15 years building Wireless Networks in Colorado. Led the sale of biz in 2024, spent 18 months merging 5 companies, now I solve SMB problems with AI.
Agent Skills by Joshua Nowak
Showing 120 vetted skills indexed across 1 GitHub repositories.
JYeswakchurn-watch
Scores account sets for churn risk with per-account evidence links and tiered rankings.
JYeswakgranola-extract
Extract decisions and action items from meeting notes with source-line citations.
JYeswakgads-review
Review Google Ads campaigns read-only with per-campaign metrics, windows, and anomaly flags.
JYeswaksecondpool-reconcile
Reconciles contradictory usage-pool wordings against quoted client reads.
JYeswakresearch-desk
Answers one hard question with three quoted sources and a falsifier.
JYeswakcitation-verify
Verifies brief citations claim by claim into a hit-or-miss ledger.
JYeswakgithub-read
Read GitHub repositories, issues, and pull requests through the github-mcp server.
JYeswakcalendar-read
Reads free/busy status and upcoming events from Google Calendar without modifying any events.
JYeswakcommunity-scan
Answers one research question from the last 30 days using read-only X and Reddit connectors.
JYeswakllm-ocr
Transcribe image-only documents with scored multi-pass OCR and quality gating.
JYeswakeval-charter
Audits a Bot's recent runs against its charter never-list and reports every breach.
JYeswakbrowser-fallback
Routes tasks between connected plugins and the Bot's browser based on probe results.
JYeswakinjection-drill
Tests Bot prompt-injection boundaries with synthetic probes and logs held or broke findings.
JYeswakaccount-health-watch
Rank usage-health and support-signal account watch lists with cited evidence per row.
JYeswakcapability-delta-review
Reviews hashed capability gate diffs to classify entitlement changes as adopt, ignore, or not applicable.
JYeswakmodel-guided-research
Simulates falsifiable hypotheses against kept run logs before changing a live Grok Bot.
JYeswakscenario-sim
Run parameterized scenarios through a local simulation and report traceable logged outputs.
JYeswaksecret-card-setup
Guides placement of API keys into masked plugin variable cards or Bot secret cards.
JYeswakmorning-briefing
Generates one dated morning briefing line summarizing carried items, overnight changes, and the first action.
JYeswakfirst-reply-desk
Answers a customer support question with one dated line using existing plugins.
JYeswakgalaxy-engineering
Posts one dated line listing pull requests waiting on a human reviewer.
JYeswakyoutube-digest
Summarize YouTube videos into timestamped briefs backed by transcript quotes.
JYeswaktalent-screen
Score and rank candidate resumes against a weighted rubric with quoted evidence per criterion.
JYeswakinbox-triage
Sorts unread email batches into act-now, draft-reply, delegate, and archive lanes with reply drafts.
Frequently Asked Questions About Joshua Nowak
FAQPage SchemaWhat tasks can I automate using JYeswak's skills?▼
You can run read-only audits and briefs across SMB operations: churn-risk scoring, Salesforce and HubSpot hygiene reports, QuickBooks book reviews, expense audits, Google Ads and paid-media reviews, inbox and Zendesk triage, meeting transcription, and cited web research. Every skill drafts, ranks, or flags with evidence; none sends, pays, posts, or edits records.
Who are these skills built for?▼
They target SMB owners, operators, and lean teams running sales, finance, marketing, and support on a Grok Bot cloud computer. The author's background is 15 years in wireless networks plus post-acquisition SMB consulting, so the skills emphasize practical revenue operations, account health, and back-office review rather than enterprise engineering.
How do the skills run in practice?▼
Each skill is a folder with native frontmatter defining its trigger conditions and hard boundaries. They execute on the Grok Bot cloud computer through connected plugins (GitHub, Stripe, Calendar, Firecrawl, X, Reddit) or the built-in browser, stopping at any login, 2FA, CAPTCHA, send, or payment step that requires a human.
Are these skills safe to run against live business data?▼
Yes, by design. Nearly every skill is explicitly read-only: it reads, ranks, cites, and drafts but never writes, sends, deletes, refunds, or publishes. Write-gated skills like zendesk-triage and sfdc-hygiene require the owner's explicit yes before any change, and secret-card-setup keeps credentials out of chat, code, and logs.
What prerequisites or dependencies do the skills need?▼
Skills depend on the Grok Bot environment and whichever plugin surface they name — for example github-mcp for repo reads, stripe-mcp for balance reports, the Google Calendar plugin for free/busy reads, or the Firecrawl plugin for public page fetches. plugin-enable-verify confirms a connector is usable before any dependent job runs.