[mybelovedstar123]

Priority

Level: low
Reason: room_ops read declared order: newest_first while returning the gateway's reply-chain order, and a small limit then kept an old replied-to message instead of newer ones. That matters to every "has anyone replied?" probe that reads a few messages.

Closes

Refs #6176. The reordering itself comes from the gateway's /messages answer, which isn't in this repo. This makes the engine's reader return the order it declares, whatever the gateway sends.

Summary

read_room (skills/agent-room-ops/read.py) took the gateway's message order as-is and labelled it newest_first. The gateway pulls a replied-to message (and a referenced one) up behind its reply, so:

• the declared order was false;
• messages[:limit] kept the replied-to message and dropped newer ones.

Now, before the limit, _newest_first() sorts by ts, newest first:

• Ties keep the gateway's order. Python's sort is stable, and that holds with reverse=True too.
• Only when every message has a numeric ts. A page where any ts is missing or non-numeric keeps the gateway's order, rather than guessing where those messages go.
• oldest_first still reverses only the rendering, now of the sorted list.

File: skills/agent-room-ops/read.py.

Checklist

• Confirmed no other open PR closes the same issue
• Git author email matches my CLA-signed email
• Single concern per PR (the read's declared order)
• Confirmed on upstream/main ("before" below)
• Tests added: section 5b of tests/room-ops-read-order.test.py (5 checks)
• Before/after evidence pasted below
• Every claim in this PR body is verifiable from the diff or the pasted output
• Live path: a read of a live AG2 Space room with a reply chain. Not run; the test drives the real read_room with the gateway's reply-chain order as the fake's answer.
• Stacked PR: N/A
• No host-specific paths added
• No CI workflow changes

Before / after evidence

The gateway answer used: reply(400, in_reply_to $1), replied-to(100), between-2(300), between-1(200).

main:

FAIL a reply chain is read newest first by timestamp — got ['reply', 'replied-to', 'between-2', 'between-1']
FAIL ...and the limit keeps the newest messages, not the replied-to one — got ['reply', 'replied-to']
FAIL ...and oldest_first is the same, reversed — got ['between-1', 'between-2', 'replied-to', 'reply']
ok   equal timestamps keep the gateway's order
ok   CONTROL: with a message lacking a numeric ts, the gateway's order is kept, not guessed
3 failure(s)

This branch: every check in room-ops-read-order passes, sections 1–6.

Test plan

• Linux (WSL2, Python 3.14): every test that reads agent-room-ops/read.py, read_room or room_ops (41 test files) passes, except two that this change doesn't touch: gateway-credential-sources (fails identically on main here) and hermetic-env-sentinels (hit my 300 s per-file timeout).
• Static CI checks: ruff 0.15.22, check-python39-compat.py on Python 3.9.25, every lint-*.sh --diff against upstream/main, gen-src-map.py --check, agents-md-sync.sh --check and review-checks.sh. All pass.

Author: discord username :: slovertop :: agent-assisted (Claude)

[github-actions]

Coverage Gate

✅ Diff coverage PASSES the 95% bar. Whole-tree (informational): 92%.

Diff Coverage

Diff: origin/main...HEAD, staged and unstaged changes

• skills/agent-room-ops/read.py (100%)

Summary

• Total: 5 lines
• Missing: 0 lines
• Coverage: 100%

[mybelovedstar123]

@cla-assistant check

(agent-assisted, Claude for slovertop)