An experimental specification and offline evaluation kit for delegated coding work: evidence schemas, a gate protocol for PASS and BLOCKED reports, and a Jev integration contract.
User repository on GitHub
An experimental protocol for evidence-aware agent handoffs, bounded worker continuation, and TypeSafe/Jev-assisted review, with reproducible evaluation.
Repo stats from the GitHub API, cached Oct 2026.
Agent Handoff Gate is an experimental protocol for delegated coding work. It checks the evidence behind a worker's PASS or BLOCKED report before the result reaches the lead agent, and describes how a controller can authorize one bounded continuation without replacing the lead's final review.
Quoted from the project's README. Read the full README.
Watch out
The measured evidence is thin and the project says so. The historical RelayLedger-40 report records two Jev runs with 38/40 and 37/40 raw labels; at a preregistered top_p threshold of 0.80 each retained 22 decisive assessments with no errors among those on that packet, and a native evaluator run scored 40/40 raw accuracy. The README calls these small-sample observations rather than production reliability or cost-saving claims, notes one baseline reported missing input sections and that both baseline semantic hashes were unresolved, and states the public result files are transcribed from the supplied report with no original API responses included. The English benchmark is a translated edition with a new identity and new hashes, and the historical measurements were not obtained on it. The 0.80 threshold is an experimental example, not a universal safety threshold and not TypeSafe's confidence field. It is an independent project, not an official TypeSafe integration. Copy reviewed against the repository README on 2026-10-04, not from running it.
skeptrunedev/jev-recruiterPython
win4r/jev-skill-suggesterPython
PyModel/jev-judge-mcpPythonPosts from builders shipping with Jev right now.
