Skills, calibrated
Where I am strong, where I am working, and where I am still learning, with the evidence on this site.
1Skills
My own rating, with the evidence on this site.
| Skill | Level | Evidence |
|---|---|---|
| PAN-OS and Panorama domain (device groups, HA, Log Collectors, XML API) | Strong | My day job as a Strata Resident Engineer; the AI-zone design |
| Directing AI coding agents with spec-first, adversarially reviewed workflows | Strong | The hub; the node script and its three reviews |
| Calibrating AI output: measure, refute, keep a feature off when it misses | Strong | The truing-up table, the router evaluation, RL name tags kept off |
| Agentic system design: routing tiers, MCP facades, claim-and-lease queue | Working | The hub; router evaluation still provisional |
| Security controls for AI agents | Working | Token and guard tests passed; hardening backlog open |
| Change safety: dry run, validate before commit, rollback, explicit yes | Working | A rule that came out of a real incident |
| PAN-OS policy and segmentation design | Working | AI-zone design, not yet on the box; offline set-syntax checker |
| LLM evaluation and local inference | Working | Large-model test job, router evaluation |
| Linux provisioning and workflow-as-code (n8n) | Working | Node script, hub generators, rollback snapshots |
| Proving tools in real environments | Learning | Several tools have run only offline or on mock data; the AI zone is not committed |
| Measuring outcomes | Learning | No hours-saved or defects-avoided figures yet |
| Hand-written code depth | Learning | My role is spec, review and verification |