Having caught one fabricated card, the obvious question is what else? So I traced every project card back to authored source material to see which ones have roots and which ones don’t.
A card is anchored if it traces to an authored, upstream source — primarily the master résumé, which is the declared source of truth (“every metric and achievement lives here first”). Two things explicitly do not count as corroboration, because both are generated from the site:
index.html.”Counting either as evidence is what let the IaC card pass the first time. So the only vote that counts is the résumé — or, for a forward-looking lab project, a real skill/experience it plainly extends.
All nine “Selected Work” cards trace cleanly to résumé bullets (Golden Helix and Zoot). No inventions in this section.
| Card | Status | Anchor in the résumé |
|---|---|---|
| Enterprise NAS migration | real | “deployed an enterprise NAS appliance with zero-data-loss cutover” — has a linked case study |
| IaC standardization | real | “Standardized AWS infrastructure with OpenTofu… automated Linux… with Ansible” |
| Company-wide MFA rollout | real | “introduced MFA and centralized Linux authentication across all user accounts” |
| Three-month ISP migration | real | “coordinated a 3-month ISP migration from locate through activation” |
| Hypervisor cluster expansion | real | “racked and integrated a new hypervisor node into the cluster” (VMware → Proxmox) |
| One-person MarTech function | real | “Assumed full ownership of marketing operations following staff departure” |
| Follow-the-Sun, Frankfurt | real | “two-year assignment in Europe to develop and launch Zoot’s Follow the Sun support model” |
| Emergency DC shutdown | real | “unplanned emergency shutdown of Zoot’s Frankfurt datacenter… turndown and recovery” |
| Cross-team protocols | real | “cutting escalations from 2 calls per 48 hours to 2 calls per month” |
This is the section the fabrication lived in. Three tiers came out of it.
| Card | Verdict | What the trace found |
|---|---|---|
| NetBox + MCP | real | Directly in the résumé: “Architected an agentic IT management system integrating NetBox with Claude via MCP.” Done, not planned. |
| This website as a case study | real | The site demonstrably exists; résumé has “Implemented deployment pipeline for web development.” |
| Home-lab reference stack | grounded plan | No past-work claim, but built on real tooling (Proxmox / OpenTofu / Ansible all in Skills). Honest as a forward-looking lab. |
| Observability stack on the lab | grounded plan | Extends real skills (Grafana, Prometheus, added in résumé v4). Forward-looking, not a false claim. |
| Email deliverability lab | grounded plan | Extends genuine MarTech work (email deliverability, segmentation, automation across 52K+ contacts). |
| Ticket triage w/ approval gates | promoted idea | No résumé anchor. Only authored trace is one bracketed example in an April feedback note (see below). |
| Runbook execution agent | promoted idea | Same: no résumé anchor; same April bracketed list. Not work done or scoped. |
| IaC diff review agent | promoted idea | Same seed; the card’s specifics (OpenTofu plan reviewer, stack, tagline) are invented on top of it. |
The audit did not confirm the IaC card as a pure zero-root fabrication. There is a single shallow seed. An April note — “LinkedIn Feedback & 3-Way Audit,” authored advice about positioning, not a résumé — contains this line:
So the phrase was seeded — as one of four illustrative examples of agent-on-ops work I could pilot. Three of those four (ticket triage, runbook execution, IaC diff review) were then promoted into concrete “Planned” project cards, each dressed with an invented stack and tagline. The fourth (log analysis) wasn’t.
That refines the verdict rather than reversing it: