open/voyage
1
0
Fork 0
Code Issues Pull requests Releases Activity

chore(voyage): release v5.6.0 — /trekexecute loop hardening

Browse source 
Bump plugin.json, package.json, package-lock.json, and the README badge
to 5.6.0; prepend CHANGELOG v5.6.0 entry + README "What's new" block for
the S38 loop-hardening work (machine-verifiable completion gate, bounded
recovery cap hierarchy + global budget, iterations_remaining signal,
fan-out hedge harmonization).

Additive — no breaking change. Canonical node --test stays 756 (0 fail);
version-consistency test green across all five version refs.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Kjell Tore Guttormsen 2026-06-20 22:00:34 +02:00
commit ef569c084e
5 changed files with 31 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

7
README.md
View file

@ -1,6 +1,6 @@
# trekplan — Brief, Research, Plan, Execute, Review, Continue
![Version](https://img.shields.io/badge/version-5.5.0-blue)
![Version](https://img.shields.io/badge/version-5.6.0-blue)
![License](https://img.shields.io/badge/license-MIT-green)
![Platform](https://img.shields.io/badge/platform-Claude%20Code-purple)
@ -10,6 +10,11 @@
A [Claude Code](https://docs.anthropic.com/en/docs/claude-code) plugin for deep implementation planning, multi-source research, autonomous execution, independent post-hoc review, and zero-friction multi-session resumption. Six commands, one pipeline:
> **What's new in v5.6.0 — `/trekexecute` loop hardening** — execution-loop termination is now **machine-verifiable** and recovery is **bounded by an explicit budget**:
> - **Machine-verifiable completion gate** — a session reports `completed` only when the Phase 7.5 manifest audit PASSes, the stop signal exits 0, and the success criteria are green (**Hard Rule 18**, promoted from prose to an enforced stop-signal contract anchored in the audit, not the model's self-assessment).
> - **Bounded recovery** — a global `TREKEXECUTE_MAX_RECOVERY_ITERATIONS` budget (default **25**, env-overridable) sits atop the per-step caps as a three-axis cap hierarchy (**Hard Rule 20**), with an `iterations_remaining` signal in the progress + summary JSON and a deterministic cross-check that catches a never-decremented counter.
> Additive — no breaking change. Full detail in [CHANGELOG.md](CHANGELOG.md).
>
> **What's new in v5.5.0 — brief framing enforcement (`brief_version 2.2`) + narrow wins** — `/trekbrief` now opens with a **framing declaration** (Phase 2.5): every brief records `framing: preserve | refine | replace | new-direction` *before* any prose is written, so the plan can't quietly polish a wrong premise. Three version-gated layers ship at `brief_version ≥ 2.2`:
> - **Framing field** — a required `framing` enum in frontmatter; enum-checked on any version, and a missing value at `≥ 2.2` is a blocker. Collected in Phase 2.5 before prose, non-skippable even in `--quick`.
> - **Memory alignment** — a new `brief-reviewer` dimension compares the brief's Intent/Goal + framing against operator memory and flags *explicit* contradictions (a no-op that scores N/A when no memory is supplied).