jevbooks

← All projects

Canny

qkal/Canny

Evidence ledger that blocks unsupported "done" claims from AI coding agents.

Verification & Guardrails55%Runner-up: Agent Decisions

What it is

Canny is a warden for AI coding agents. It hooks into Claude Code and Codex CLI, keeps an append-only ledger of what the agent actually did, and refuses to let the agent finish on an unsupported claim.

How it uses Jev

Jev answers two Noul yes/no questions: whether an edit breaks a rule in CLAUDE.md or AGENTS.md (a note only), and whether the agent's final message claims the work is done. The second can only relax the done-gate, never tighten it.

Primitives:noul

Technique worth stealing

Cache Jev answers by content hash and act only outside a wide confidence band, so verdicts stay stable despite model drift.

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricno0.35
  • Shows a System One patternno0.31
  • Handles uncertaintyno0.05
  • Measuredno0.05
  • Runnableyes0.96
  • Worth recommendingno0.18
  • Model replicano0.08
  • Problem scopescore on a 0–2 scale1.38
  • About Jevyes0.86

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.