embedded-target

Scaffold the embedded target loop in this project — DEVICE.md plus the harness/, faults/, and repros/ skeletons.

Drive the on-device debug loop against a remote embedded target — reproducible (native/injected/replayed) or instrument-and-wait (irreducible environment dependence). Use when debugging behavior that only appears on-device, where the limiting resource is occurrences of the triggering condition.

Synthesize environment trigger conditions (dropouts, lossy/jittery links, partitions) as commands, and sweep the dropout-duration × protocol-phase space to make a timing-dependent bug deterministic. Use when a bug "depends on timing" or on network degradation and is hard to reproduce by hand.

Build or fill in the build-deploy-run-collect harness for a remote embedded target — the scripted loop an agent drives unattended. Use when setting up a new device project or when harness commands are missing or hand-run.

Capture and analyze traffic on a video-heavy embedded link — headers-only ring buffer always on, control-plane in full, batch tshark analysis the agent reads as conclusions. Use when diagnosing protocol, loss, reordering, jitter, or timing bugs on the device's network.

Instrument an embedded target before the bug, not per-bug — structured boundary events, counters, a journal siphon that survives fast rollover, episodic capture on rate/signature triggers, and runtime log-level control without restarts. Use when setting up observability or when an occurrence was wasted because logs didn't cover the failing region.

1 hook across 1 event