agent-signals

Screen claude/quoted-dialog

A reply that quotes a captured dialog, with a rule above the quote, above a framed box. The expected state is wrong on purpose. Do not "fix" it. The version and the date are those of the dialog above a box. `tests/test_sessions.py::test_a_reply_that_quotes_a_dialog_reads_as_that_dialog` asserts the state.

Expected state
DIALOG_OVER_BOX known false positive
Agent version
Claude Code 2.1.283
Provenance
derived
The text of the screen claude/quoted-dialog, a plain capture.
● The pane of that session reads:
────────────────────────────────────────────────────────────────────────────────
  Teach auto mode about your environment?

  ❯ 1. Yes
    2. Not now
    3. Don't show again

  Enter to confirm · Esc to cancel
────────────────────────────────────────────────────────────────────────────────
❯ so the classifier answers IDLE
────────────────────────────────────────────────────────────────────────────────
  ⏵⏵ auto mode on (shift+tab to cycle) · ? for shortcuts · ← for agents

Download the exact bytes of this screen Read the metadata file on GitHub

Expected result

state
DIALOG_OVER_BOX

Metadata

agent
claude
agent_version
2.1.283
measured
2026-09-28
capture
plain
pane
not known
provenance
derived
derived_from
hand-built
known_false_positive
true
known_false_positive_reason
A reply that quotes a dialog holds every part of the rule for a dialog above a box, and no text on the screen tells the quote apart from a real dialog. A reader that gives DIALOG_OVER_BOX sends no key, which is fail-safe. A tighter rule reads a real dialog as an idle box and sends an Enter into it.
facts