Repository navigation
Possible complementary direction: LongHorizon-Harness for sustained agent tasks #508
Description
Hi, I'd like to suggest a complementary direction for loop-engineering.
Project:
https://github.com/AMAP-ML/LongHorizon-Harness
Your project focuses on coding-agent and software-engineering workflows.
LongHorizon-Harness focuses on carrying one task across many rounds without state drift. It keeps the goal and verified progress in durable state, starts each execution round with a fresh context, and uses an independent Auditor before accepting progress.
Why this may be useful here:
- A concise lifecycle mapping could compare its existing concepts with Manager, fresh-context Executor, independent Auditor, and durable verified state.
- it adds an outer task-lifecycle layer without claiming to replace the project's existing agent or runtime
If this direction fits, I can prepare a small documentation or example contribution following the repository's conventions.
I am part of the team behind LongHorizon-Harness, so I wanted to disclose that connection.
Paper:
https://arxiv.org/abs/2608.01964
Project page:
https://lh-harness.pages.dev
Activity
cobusgreyling commentedon Aug 17, 2026
Thanks for the pointer and for disclosing that you are on the LongHorizon-Harness team.
Durable STATE, a fresh-context implementer, and an independent verifier are already first-class here (
LOOP.md,docs/safety.md). Whether we add an official complementary-tools mapping to an external harness is a maintainer product call — I will not decide that in this comment.If a maintainer wants a docs/example, Path A in
CONTRIBUTING.mdis the right route (examples/or a shortdocs/note). Please wait for that signal before opening another draft (the earlier #506 draft was closed).