Evidence record / openhands
A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the issues that cite it, below.
2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-and
OpenHands ships its own agent-driven QA harness and recipes for 27 feature families and documented known bugs (v1.26.0 to v1.27.0). The repo now ships verify-openhands: a control-openhands CLI that launches an isolated real stack (agent server, automation, static frontend, ingress), drives a browser, and records evidence (screenshots, ARIA snapshots, video), plus a feature map of 27 areas (F01 to F27) with exact recipes, expected results and known failures for each user-facing behavior. v1.27.0 records documented known bugs in the map; coverage requirements are not proof of exhaustive defect coverage, and agent-written PRs in the window cite its live runs as their evidence (for example #18127 and #18039).
Channel: tagged-release (follow-ups on main only). Half: capability. Date: 2026-10-08 (v1.26.0); 2026-10-12 (v1.27.0).
Operator consequence: The map is a detailed public account of how Canvas is intended to behave, by-design edges included (the model-pill rule at F13 L23 and L85), and its recipes can be rerun against your own stack to check an upgrade instead of trusting release notes. It runs a real stack and browser (Node 24, uv, Chromium) and its skill forbids unauthorized external writes, but that instruction is not technical confinement: the local terminal runs as the user’s account. Human attention shifts from clicking through QA to reviewing agent-recorded evidence; the vendor now writes its review rules around that evidence.
Receipt
- https://github.com/OpenHands/OpenHands/commit/d7e14e25bcf14238e566139ec9490e1f05efcd69
- https://github.com/OpenHands/OpenHands/pull/17961
- https://github.com/OpenHands/OpenHands/pull/18304
- https://github.com/OpenHands/OpenHands/blob/v1.27.1/.agents/skills/verify-openhands/SKILL.md
- https://github.com/OpenHands/OpenHands/blob/v1.27.1/.agents/skills/verify-openhands/references/feature-map/F13-agent-profiles.md#L23
Finding metadata
Run: 2026-10-13-profile-refresh-2026-09-21_2026-10-13-frontier-v0
Finding ID: 2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-and
Profile citations
- OpenHands / claim / qa-recipes-and-evidence
Source links
Primary links, including exact changelog lines when available.
- commit2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-andgithub.com/OpenHands/OpenHands/commit/d7e14e25bcf14238e566139ec9490e1f05efcd69merged pr2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-andgithub.com/OpenHands/OpenHands/pull/17961merged pr2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-andgithub.com/OpenHands/OpenHands/pull/18304tagged commit file2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-andOpenHands/OpenHands / .agents/skills/verify-openhands/SKILL.mdtagged commit file2026-10-13-openhands-openhands-ships-its-own-agent-driven-qa-harness-and-recipes-for-27-feature-families-andOpenHands/OpenHands / .agents/skills/verify-openhands/references/feature-map/F13-agent-profiles.md#L23
Versioned source: run artifact