Could augment it with a big all-caps prohibition in AGENTS.md…

OK, let’s try

@nutritious-balloon-65098 find the workflow that posts a comment about “HUMAN:” and understand its purpose. Then make it a CI step instead, and warn in AGENTS.md that it’s human-only and AI agents MUST NOT touch it. Also, since we’re at it, this workflow should check that the PR template was respected; if not, fail too.

Since my last summary, there have been no additional code changes or tool actions. Checklist: • ✅ The requested workflow change was completed and committed locally. • ✅ The HUMAN reminder behavior was converted into a failing CI-style PR description check. • ✅ PR template compliance checks were added. • ✅

AGENTS.md

now clearly warns that the

HUMAN:

section and human-tested checkbox are human-only and AI agents must not touch them. • ✅ Targeted tests and pre-commit checks passed before commit. • ✅ Changes were committed with the required co-author trailer. • ✅ No push or PR was created, following the instruction not to do so without explicit confirmation. Conciseness: • The changes are focused on the requested PR description/HUMAN validation workflow and supporting tests. • I do not see extraneous changes that need reverting. Current state: • Local commit:

6676ab96 Add PR description readiness check

• Awaiting your preference on whether to push/create a PR or keep changes local.

Since my last summary, I only answered the follow-up question: • The HUMAN reminder workflow was fully implemented in

OpenHands/OpenHands

as

.github/workflows/pr-readiness-confirm.yml

. • It was not a reusable workflow or custom reusable action. • It used

actions/github-script@v9

as an execution helper, but the actual HUMAN-check logic was inline JavaScript embedded directly in the workflow YAML. •

software-agent-sdk

did not already have that workflow; before my changes, it only had the human-tested checkbox in the PR template.

OpenHands encountered an error: **Failed to send message to agent server: HTTP 503 error: no available server ** See the conversation for more information.

Since my last summary: • ✅ Re-checked the PR after you noted both you and the AI reviewer had left comments. • ✅ Found and addressed your human review comment: ◦ Removed the unnecessary

synchronize

trigger from

.github/workflows/pr-description-check.yml

. ◦ Replied to the review thread and resolved it. • ✅ Addressed the naming clarity concern: ◦ Renamed

REQUIRED_SECTIONS

to

TEMPLATE_SECTIONS

. ◦ Renamed

REQUIRED_FILLED_SECTIONS

to

NONEMPTY_SECTIONS

. • ✅ Addressed the latest AI reviewer feedback where useful: ◦ Added concise workflow comments explaining the `pull_request_target`/base-SHA security pattern. ◦ Kept comments minimal and directly relevant. • ✅ Validated and pushed the final changes in commit

36a15950

. • ✅ Confirmed the PR is non-draft, all review threads are resolved, and CI is green: ◦

33

passed ◦

0

failed ◦

0

pending ◦ latest QA report: PASS Conciseness check: • The latest changes are focused on review feedback only. • No extraneous files or unrelated changes were added. • I did not modify the human-only

HUMAN:

section or human-tested checkbox. Remaining status: • PR #3527 is mergeable but still blocked by the repository’s required human approval/review policy.

Since the last summary, I completed the remaining review/iteration work on PR #3527. Checklist: • ✅ Simplified validation as requested: ◦ Only

REQUIRED_TEMPLATE_FIELDS = ("Why", "Summary", "How to Test")

are enforced. ◦ Those three fields must be present and nonempty. ◦ Optional sections like

Issue Number

,

Video/Screenshots

,

Type

, and

Notes

may now be deleted by a human. • ✅ Applied the latest AI reviewer suggestions: ◦ Added concise documentation for the human-note length rationale. ◦ Added a small docstring for

visible_text()

. • ✅ Updated tests to match the new intended behavior. • ✅ Updated the PR description outside the human-only block to remove the stale claim about Type checkbox enforcement. • ✅ Replied to and resolved the AI QA review thread about Type checkbox enforcement. • ✅ Requested re-review. • ✅ Verified the PR is now green: ◦ All GitHub checks pass. ◦ No unresolved review threads. ◦ AI review is approved. ◦ Latest QA is “PARTIAL” only because the newly added

pull_request_target

workflow cannot run live until it exists on `main`; no functional issue was found. ◦ PR is clean/mergeable. Conciseness: • The changes are focused on the requested validation scope and reviewer feedback. • I did not leave extraneous functional changes. • The only small additions are comments/docstrings directly tied to reviewer suggestions.