Erick Shepherd

← All skills

Skill

AI output verification

Related work and learning, in one place.

Checking AI-generated work and procedural claims against tests, execution records, and other evidence. Verification can use deterministic tools as well as human review.

Projects

  • excubitor — Safety Fences for Autonomous Coding Agents

    Project

    I built excubitor to enforce Git checks in unattended Claude Code workflows. Python hooks block specified actions outside the model, while a separate audit checks project claims against executable evidence.

  • did-it — Checking an AI Agent's Claims Against Its Transcript

    Project

    I built did-it to compare a coding agent’s procedural claims with its Claude Code transcript. It checks the sequence of tool calls and results, then returns a receipt identifying supporting, contradictory, or missing evidence for each claim.

  • Coding-Agent Evaluation Harness

    Project

    I developed an AI-assisted Python harness to compare coding-agent results against acceptance checks fixed before each run. Recorded provider runs and test results support repeatable comparisons on the selected tasks.

Courses & credentials