Agents heat 1 · Quiet★ 0heat incomplete (no X)

plan-anchor

A MIT Claude Code/Codex skill that captures planning directives as numbered anchors, tunes typed Jev questions, and gates both the plan and delivered work against your own quoted words via hosted TypeSafe Jev.

What it is

Keep plans and code anchored to what you actually said.

What it does: Captures planning directives as anchors and judges plan beads and sprint reports with hosted Jev.

Why it belongs in the showcase: Distinct dual-gate plan-drift control on TypeSafe Jev, not another model router.

Caveats: Explicit remote consent and credential required.

Evidence

Public MIT repo, created 2026-10-10, 0 stars as of 2026-11-04. README documents TYPESAFE_API_KEY / platform credential store, allow_remote_judge consent, planning and development gates, and design docs under docs/design/.

Caveats

  • Requires allow_remote_judge:true and a TypeSafe credential before any remote Jev call
  • Sends scoped planning/delivery text to api.typesafe.ai when enabled
  • Deterministic checks still needed alongside Jev text judgments

Unofficial case page. Stars / list position observed for heat; do not treat heat as a quality SLA. Review the upstream README before you depend on it.