John Vedovino — tools & audits

n8n template: a branded weekly Meta Ads PDF (spend, CPL, ROAS, top/bottom ads, Claude summary) emailed every Monday. Proven on sample data.

CLI that flags a listing contradicting itself: digital vs. shipping, unqualified claims, undisclosed AI, Etsy limits. 0 false flags on 133 live drafts.

Forecasts an agent session's final token burn from its first 5 turns. Measured leave-one-out median error: 59.7% (a rough early warning).

Single-file scan for eval/exec, shell injection, pickle, unsafe yaml, SQL string-building, DOM sinks and secret-looking literals. No network, no key.

Screenshot-based lookalike-page detector for phishing triage.

Save, fork and replay agent sessions from checkpoints.

Explore a packet capture entirely in the browser — nothing uploaded.

A small reinforcement-learning teaching kit you can run and read in an afternoon.

A tiny on-device classifier game: persuade the gatekeeper.

Printable birdsong field-note cards.

An editor that records how a piece of writing was typed, on the writer's own machine.

Five n8n workflows for appointment businesses: reminders, no-show rescue, waitlist fill, invoice dunning, Monday KPI email.

I install the no-show pack on your own sheet and accounts. You only accept once it works on your data.

Automated WCAG audit of up to 5 pages (axe-core + Lighthouse), plain-English PDF, one re-check. Automated only, stated plainly.

Every pinned file in your uv.lock checked against PyPI's signed publishing records, its hash and its source repo.

GitHub Actions audit for cache-poisoned secret leaks: credentials in cached folders, PR jobs restoring privileged caches.

An exact exploration score for your AI agent in ten generated rule worlds. Upwork Project Catalog.

Terminal-output rendering safety review: your ANSI-to-HTML converter fuzzed in an isolated, network-off browser.

Your AI agent's session logs checked against your safety rules with a temporal-logic monitor.

Your agent's failure cut down to the turns that cause it.

MCP server configuration review: tool descriptions that steer your agent, sensitive paths, hidden characters.