[TimeToBuildBob]

Problem

ActivityWatch/activitywatch#1487 reports a Bluetooth controller keeping Windows AFK status active while untouched. Windows uses GetLastInputInfo, not the pynput listener backend; a listener-only filter would not address it. The event source remains unknown.

Scope

Adds a standalone, opt-in, standard-library-only Windows diagnostic and comparison procedure. It samples Windows idle time and aggregate mouse/keyboard injection flags without recording keys, text, coordinates, window titles, device IDs or raw event timestamps. Hooks forward events and clean up on exit and partial failure. Ctrl+C inside observation is deferred until input has been forwarded.

No production backend/config changes, blanket injected-input filtering, or controller-causality claim. Unflagged does not prove physical input; missing counts do not prove no activity. Documentation covers coverage and correlation limits.

Verification

• xvfb-run -a make test: 22 existing AFK tests + 14 diagnostic tests passed; package typecheck passed.
• mypy --platform win32 misc/windows_input_diagnostic.py --ignore-missing-imports: passed.
• ruff check and git diff --check: passed.
• Independent native-ABI/privacy/lifecycle review; repaired callback cancellation defect with a failing-then-passing test.
• Hook callback failures are contained, never escaping the ctypes boundary: classification exceptions, an interrupt raised on the pass-through call's return, and any other pass-through failure are all deferred and surfaced on the next state check, including after the final sample.
• Unrestricted unittest discovery reaches an existing live-server test requiring localhost:5666; unavailable locally. No all-suite pass claimed.

Windows execution and reporter reproduction are unverified. API doubles test lifecycle control flow, not native ABI or hook delivery. The connected-idle/disconnected-idle procedure in docs/windows-input-diagnostic.md is the next evidence gate.

Related to ActivityWatch/activitywatch#1487; does not close the production issue.

[greptile-apps]

Confidence Score: 5/5

[Low risk] Adds a read-only Windows diagnostic tool and documentation.

The PR appears safe to merge based on the reviewed changes.

Summary

This PR adds an opt-in Windows diagnostic that compares idle time with aggregate keyboard and mouse injection flags without changing AFK behavior.

• Documents a connected-versus-disconnected controller comparison and the limits of what hook counts can establish.
• Adds lifecycle and event-aggregation tests to the default test target.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  W[Windows idle API] --> S[Interval sample]
  H[Keyboard and mouse hooks] --> C[Aggregate flag counts]
  C --> S
  S --> J[JSONL diagnostic output]
  J --> U[User compares idle periods]

Loading

_{Reviews (3) · Last reviewed commit: "test(diagnostic): drive message loop wit..."}

[TimeToBuildBob]

CI-green and mergeable — waiting only on a maintainer click.

This PR is ready to merge, but the bot has pull-only access to this repo and can't self-merge — surfacing it here so it isn't lost. The monitoring loop will stop re-flagging it now that this note is posted.

[TimeToBuildBob]

@greptileai review

[TimeToBuildBob]

🤖 AI code review

Adds a standalone Windows diagnostic that samples GetLastInputInfo and counts low-level mouse and keyboard events by injection flags. Includes hook lifecycle handling, JSON output, a comparison procedure, and unit tests with native API doubles. The production AFK backend is unchanged.

Safe to merge — no P0/P1 findings

Confidence 5/5

✅ No findings. The diff looks correct to me on this pass.

Files changed (5) — the diff as I read it

• Makefile — Adds discovery of test_windows_input_diagnostic.py to the test target.
• README.md — Adds a Windows unexpected input diagnostic section linking to the comparison documentation.
• docs/windows-input-diagnostic.md — Documents execution, controller comparison steps, JSON fields, injection flag interpretation, and observation limits.
• misc/windows_input_diagnostic.py — Adds flag classification, interval counters, idle sampling, Windows hook installation and cleanup, deferred callback errors, and a duration-bounded CLI.
• tests/test_windows_input_diagnostic.py — Adds tests for classification, wraparound, interval aggregation, hook cleanup, callback failures, cancellation, and event delivery using native API doubles.

Previous review passes

commit	score	findings	engine	when
8a10896c647f	4/5	1	subscription	2026-10-03 20:23 UTC
d3f87a063a28	4/5	1	subscription	2026-10-03 21:24 UTC
a202201ab311	3/5	3	llm	2026-10-03 23:22 UTC
08305cf8d281	5/5	0	subscription	2026-10-04 00:28 UTC
08305cf8d281	5/5	0	subscription	2026-10-09 06:52 UTC
08305cf8d281	5/5	0	subscription	2026-10-09 07:27 UTC
08305cf8d281	5/5	0	subscription	2026-10-09 08:21 UTC
08305cf8d281	4/5	1	llm	2026-10-09 09:03 UTC

_{Reviewed 08305cf8d281 · panel: codex:gpt-6.1-sol · subscription engine · 67s · about this reviewer}

_{Maintainer commands}

_{@TimeToBuildBob review (own line) — fresh review · @TimeToBuildBob fix — a worker acts on the findings. Once per comment; 👀 = received.}

[TimeToBuildBob]

@greptileai review