kevin/meshtastic_firmware
1
0
Fork 0
Code Issues Pull Requests Actions 4 Packages Projects Releases Wiki Activity

Add MCP server for interacting with meshtastic devices and testing framework / TUI (#10194)

Browse Source 
* Start of MCP server and test suite

* Add MCP server for interacting with meshtastic devices and testing framework / TUI

* Update mcp-server/README.md

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* fix mcp-server review feedback from thread

Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/91dc128a-ed50-4d07-8bb2-3dc6623a05f7

Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com>

* Enhance StreamAPI and PhoneAPI for improved log record handling and concurrency control

* Semgrep fixes

* Trunk and semgrep fixes

* optimize pio streaming tee file writes

Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/04e26c6b-6a2b-45be-bbeb-79ae4d0be633

Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com>

* chore: remove redundant log handle assignment

Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/04e26c6b-6a2b-45be-bbeb-79ae4d0be633

Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com>

* Consolidate type imports and remove placeholder test files

* Add tests for config persistence and more exchange messages

* Refactor position test to validate on-demand request/reply behavior

* Remove  position request/reply test and update README for telemetry behavior

* Fix transmit history file to get removed on factory reset

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
This commit is contained in:
Ben Meadors
2026-04-18 11:29:02 -05:00
co-authored by Copilot copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
parent 8fd0a7f283
commit 6b15571e14
77 changed files with 10701 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/copilot-instructions.md
+160
View File
@@ -429,6 +429,8 @@ Most workflows can be triggered manually via `workflow_dispatch` for testing.
## Testing
### Native unit tests (C++)
Unit tests in `test/` directory with 12 test suites:
- `test_crypto/` - Cryptography
@@ -446,6 +448,164 @@ Run with: `pio test -e native`
Simulation testing: `bin/test-simulator.sh`
### Hardware-in-the-loop tests (`mcp-server/tests/`)
Separate pytest suite that exercises real USB-connected Meshtastic devices. See the **MCP Server & Hardware Test Harness** section below for invocation, tier layout, and agent usage rules.
## MCP Server & Hardware Test Harness
The `mcp-server/` directory houses a firmware-aware [MCP](https://modelcontextprotocol.io/) server plus a pytest-based integration suite. AI agents that speak MCP get a well-defined tool surface for flashing, configuring, and inspecting physical Meshtastic devices — use it instead of hand-rolling `pio` or `meshtastic --port` calls where possible. `mcp-server/README.md` is the operator-facing setup doc; this section is the agent-facing usage contract.
The repo registers the server via `.mcp.json` at the repo root — Claude Code picks it up automatically once `mcp-server/.venv/` is built (`cd mcp-server && python3 -m venv .venv && .venv/bin/pip install -e '.[test]'`).
### When to use which surface
| Goal | Tool |
| ------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| Find a connected device | `mcp__meshtastic__list_devices` |
| Read a live node's config/state | `mcp__meshtastic__device_info`, `list_nodes`, `get_config` |
| Mutate a device (owner, region, channels, reboot) | `set_owner`, `set_config`, `set_channel_url`, `reboot`, `shutdown`, `factory_reset` — all require `confirm=True` |
| Flash firmware to a variant | `pio_flash` (any arch) or `erase_and_flash` (ESP32 factory install) |
| Stream serial logs while debugging | `serial_open` → `serial_read` loop → `serial_close` |
| Administer `userPrefs.jsonc` build-time constants | `userprefs_get`, `userprefs_set`, `userprefs_reset`, `userprefs_manifest` |
| Run the regression suite | `./mcp-server/run-tests.sh` (or `/test` slash command) |
| Diagnose a specific device | `/diagnose [role]` slash command (read-only) |
| Triage a flaky test | `/repro <node-id> [count]` slash command |
**One MCP call per port at a time.** `SerialInterface` holds an exclusive OS-level lock on the serial port for its lifetime. If a `serial_*` session is open on `/dev/cu.usbmodem101`, calling `device_info` on the same port will fail fast pointing at the active session. Sequence calls: open → read/mutate → close, then next device. Never parallelize tool calls on the same port.
### MCP tool surface (~32 tools)
Grouped by purpose. Full argument shapes in `mcp-server/README.md`; a few high-value signatures are called out here.
- **Discovery & metadata**: `list_devices`, `list_boards`, `get_board`
- **Build & flash**: `build`, `clean`, `pio_flash`, `erase_and_flash` (ESP32 only), `update_flash` (ESP32 OTA), `touch_1200bps`
- **Serial sessions** (long-running, 10k-line ring buffer): `serial_open`, `serial_read`, `serial_list`, `serial_close`
- **Device reads**: `device_info`, `list_nodes`
- **Device writes** (all require `confirm=True`): `set_owner`, `get_config`, `set_config`, `get_channel_url`, `set_channel_url`, `send_text`, `reboot`, `shutdown`, `factory_reset`, `set_debug_log_api`
- **userPrefs admin** (build-time constants, not runtime config): `userprefs_get`, `userprefs_set`, `userprefs_reset`, `userprefs_manifest`, `userprefs_testing_profile`
- **Vendor escape hatches**: `esptool_chip_info`, `esptool_erase_flash`, `esptool_raw`, `nrfutil_dfu`, `nrfutil_raw`, `picotool_info`, `picotool_load`, `picotool_raw`
`confirm=True` is a tool-level gate on top of whatever permission prompt your MCP host shows. **Don't bypass it** by asking the host to auto-approve — it exists specifically because MCP hosts sometimes remember "always allow this tool" and that's dangerous for `factory_reset` and `erase_and_flash`.
### Hardware test suite (`mcp-server/run-tests.sh`)
The wrapper auto-detects connected devices (VID → role map: `0x239A` → `nrf52`, `0x303A`/`0x10C4` → `esp32s3`), maps each role to a PlatformIO env (`nrf52` → `rak4631`, `esp32s3` → `heltec-v3`, overridable via `MESHTASTIC_MCP_ENV_<ROLE>`), then invokes pytest. Zero pre-flight config needed from the operator.
Suite tiers (collected + run in this order via `pytest_collection_modifyitems`):
1. `tests/unit/` — pure Python (boards parse, pio wrapper, userPrefs parse, testing profile). No hardware.
2. `tests/test_00_bake.py` — flashes each detected device with current `userPrefs.jsonc` merged with the session's test profile. Has its own skip-if-already-baked check comparing region + primary channel to the session profile; skips cheaply on warm devices.
3. `tests/mesh/` — multi-device mesh: bidirectional send, broadcast delivery, direct-with-ACK, mesh formation within 60s. Parametrized `[nrf52->esp32s3]` and `[esp32s3->nrf52]`.
4. `tests/telemetry/` — `DEVICE_METRICS_APP` broadcast timing.
5. `tests/monitor/` — boot-log panic check.
6. `tests/fleet/` — PSK seed session isolation.
7. `tests/admin/` — channel URL roundtrip, owner persistence across reboot.
8. `tests/provisioning/` — region + modem + slot bake, admin key presence, `UNSET` region blocks TX, userPrefs survive factory reset.
Invocation patterns:
```bash
./mcp-server/run-tests.sh # full suite (auto-bake-if-needed)
./mcp-server/run-tests.sh --force-bake # reflash before testing
./mcp-server/run-tests.sh --assume-baked # skip bake (caller vouches for device state)
./mcp-server/run-tests.sh tests/mesh # one tier
./mcp-server/run-tests.sh tests/mesh/test_direct_with_ack.py # one file
./mcp-server/run-tests.sh -k telemetry # name filter
```
**No hardware detected?** The wrapper auto-narrows to `tests/unit/` only and prints `detected hub : (none)` in the pre-flight header. Agents interpreting the output should call this out explicitly — a 52-test green run without hardware is qualitatively different from a 12-unit-test green run.
**Artifacts every run produces:**
- `mcp-server/tests/report.html` — self-contained pytest-html. Each test gets a `Meshtastic debug` section with the tail of firmware log + device state dump. **Open this first** on failures; it's the canonical evidence source.
- `mcp-server/tests/junit.xml` — CI-parseable.
- `mcp-server/tests/reportlog.jsonl` — pytest-reportlog stream (`$report_type` keyed JSONL). Consumed by the live TUI.
- `mcp-server/tests/fwlog.jsonl` — firmware log mirror from the `meshtastic.log.line` pubsub topic. Populated by the `_firmware_log_stream` autouse session fixture.
### Live TUI (`meshtastic-mcp-test-tui`)
A Textual-based live view that wraps `run-tests.sh`. Tails reportlog for per-test state, streams firmware logs, polls device state at startup + post-run (gated out of the active run because `hub_devices` holds exclusive port locks). Key bindings:
| Key | Action |
| --- | ------------------------------------------------------------------------------------------------------------ |
| `r` | re-run focused test (leaf → that node id; internal node → directory or `-k`) |
| `f` | filter tree by substring |
| `d` | failure detail modal (pulls `longrepr` + captured stdout from the reportlog) |
| `g` | export reproducer bundle (tar.gz with README, test_report.json, time-filtered fwlog, devices.json, env.json) |
| `l` | toggle firmware log pane |
| `x` | tool coverage modal |
| `c` | cross-run history sparkline |
| `q` | quit (SIGINT → SIGTERM → SIGKILL escalation, 5-s windows each) |
Launch:
```bash
cd mcp-server
.venv/bin/meshtastic-mcp-test-tui # full suite
.venv/bin/meshtastic-mcp-test-tui tests/mesh # args pass through to pytest
```
The plain CLI stays primary; the TUI is for operators who want a live dashboard. Both consume the same `run-tests.sh`.
### Slash commands (Claude Code + Copilot)
Three AI-assisted workflows wrap the test harness. Claude Code operators get `/test`, `/diagnose`, `/repro`; Copilot operators get `/mcp-test`, `/mcp-diagnose`, `/mcp-repro`. Bodies:
- `.claude/commands/{test,diagnose,repro}.md`
- `.github/prompts/mcp-{test,diagnose,repro}.prompt.md`
`.claude/commands/README.md` is the index.
House rules for agents running these prompts:
- **Interpret failures, don't just echo them.** Pull firmware log tails from `report.html` and classify each failure as transient / environmental / regression. Use the exact format in `.claude/commands/test.md`.
- **No destructive writes without operator approval.** Any skill that could reflash, factory-reset, or reboot a device must describe the action and stop. The operator authorizes.
- **Sequential MCP calls per port.** See above.
- **"Unknown" is a valid classification.** If evidence doesn't support a root cause, say so and list what would disambiguate. Do not invent.
### Key fixtures (test authors + agents debugging)
`mcp-server/tests/conftest.py` provides:
- **`_session_userprefs`** (autouse session) — snapshots `userPrefs.jsonc` at session start, merges the session test profile via `userprefs.merge_active(test_profile)`, restores at teardown. Four layers of safety: pytest teardown + `atexit` + sidecar file (`userPrefs.jsonc.mcp-session-bak`) + startup self-heal in `run-tests.sh`. **Do not edit `userPrefs.jsonc` from inside a test.**
- **`_firmware_log_stream`** (autouse session) — subscribes to `meshtastic.log.line` pubsub on every connected `SerialInterface` and mirrors lines to `tests/fwlog.jsonl`. Drives the TUI firmware-log pane.
- **`_debug_log_buffer`** (autouse per-test) — captures last 200 firmware log lines + device state for attachment to the pytest-html `Meshtastic debug` section on failure.
- **`hub_devices`** (session) — `dict[role, SerialInterface]` with session-long exclusive port locks. Reason the TUI's device poller is gated to startup + post-run only.
- **`baked_mesh`** — parametrized mesh-pair fixture; depends on `test_00_bake`. `pytest_generate_tests` in `conftest.py` auto-generates `[nrf52->esp32s3]` and `[esp32s3->nrf52]` variants.
- **`test_profile`** — session-scoped dict: region, primary channel, admin key, PSK seed. Derived from `MESHTASTIC_MCP_SEED` (defaults to `mcp-<user>-<host>`).
### Firmware integration points tied to the test harness
Two firmware changes exist specifically so the test harness works reliably. **Keep these in mind when touching related code.**
- **`src/mesh/StreamAPI.cpp` + `StreamAPI.h`** — `emitLogRecord` uses a dedicated `fromRadioScratchLog` + `txBufLog` pair and a `concurrency::Lock streamLock`. Before this fix, `debug_log_api_enabled=true` would tear `FromRadio` protobufs on the serial transport because `emitTxBuffer` and `emitLogRecord` shared a single scratch buffer. The conftest enables the log stream session-wide; without this fix the device would corrupt its own FromRadio replies mid-session.
- **`src/mesh/PhoneAPI.cpp`** — `ToRadio` `Heartbeat(nonce=1)` triggers `nodeInfoModule->sendOurNodeInfo(NODENUM_BROADCAST, true, 0, true)` for serial clients, mirroring the pre-existing behavior for TCP/UDP clients in `PacketAPI.cpp`. The mesh tests rely on this to force a NodeInfo broadcast right after connect so the peer discovers them before the test's first assertion.
If you're modifying `StreamAPI`, `PhoneAPI`, `NodeInfoModule`, or `userPrefs` flow, run `./mcp-server/run-tests.sh` at minimum before asking for review.
### Recovery playbooks
| Symptom | First check | Fix |
| ---------------------------------------------------------- | ------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `userPrefs.jsonc` dirty after test run | `git status --porcelain userPrefs.jsonc` | If non-empty, re-run `./mcp-server/run-tests.sh` once — the pre-flight self-heal restores from sidecar. If still dirty, `git checkout userPrefs.jsonc`. |
| Port busy / wedged CP2102 on macOS | `lsof /dev/cu.usbserial-0001` | Kill the holder. USB replug if the kernel still reports busy. Often a stale `pio device monitor` or zombie `meshtastic_mcp` process. |
| nRF52 appears unresponsive | `list_devices` shows VID `0x239A` but `device_info` times out | `touch_1200bps(port=...)` drops it into the DFU bootloader → `pio_flash` re-installs. |
| Multiple MCP server processes | `ps aux \| grep meshtastic_mcp` shows >1 | Kill all but the one your MCP host spawned. Zombies hold ports and break tests. |
| Mesh formation fails, one side sees peer but other doesn't | `/diagnose` (or `list_nodes` on both sides) | Asymmetric NodeInfo. `test_direct_with_ack` has a heal path; `/repro` it a few times. If persistent, both devices' clocks may be out of sync with their NodeInfo cooldown. |
| "role not present on hub" in skip reasons | `list_devices` | Expected if a device is unplugged. Reconnect before re-running the tier. |
| Tests fail only on first attempt then pass on rerun | — | State leak from a prior session. Run with `--force-bake` to reset to a known state. |
### Never do these without asking
- `factory_reset` — wipes node identity; regenerates PKI keypair. Mesh peers will reject old DMs until re-exchange. Legitimate only when the operator explicitly wants it.
- `erase_and_flash` — full chip erase; destroys all on-device state.
- `esptool_erase_flash` / `esptool_raw` write/erase — bypasses pio's safety chain.
- `set_config` on `lora.region` — changes regulatory domain; requires physical-location context the operator has and the agent doesn't.
- `reboot` / `shutdown` mid-test — breaks fixture invariants.
- `push -f`, `rebase -i`, `reset --hard`, or any history-rewriting git operation.
- Clicking computer-use tools on web links in Mail/Messages/PDFs — open URLs via the claude-in-chrome MCP so the extension's link-safety checks apply.
## Resources
- [Documentation](https://meshtastic.org/docs/)
.github/prompts/mcp-diagnose.prompt.md
+57
View File
@@ -0,0 +1,57 @@
---
mode: agent
description: Device health report via the meshtastic MCP tools (Copilot equivalent of the Claude Code /diagnose slash command)
---
# `/mcp-diagnose` — device health report
Equivalent of `.claude/commands/diagnose.md`. Use when the operator asks to "check the devices", "what's the mesh looking like", "is nrf52 alive", etc.
This prompt assumes the meshtastic MCP server is registered with your VS Code Copilot agent. If it isn't, fall back to running `./mcp-server/run-tests.sh tests/unit` plus a short `device_info` script via the terminal.
## What to do
1. **Enumerate hardware** via the `list_devices` MCP tool (with `include_unknown=True`). For each entry where `likely_meshtastic=True`, capture `port`, `vid`, `pid`, `description`.
2. **Apply the operator's filter** (if any):
- No filter → every likely-meshtastic device.
- `nrf52` → `vid == 0x239a`
- `esp32s3` → `vid == 0x303a` or `vid == 0x10c4`
- A `/dev/cu.*` path → only that port.
- Anything else → substring match on port.
3. **For each selected device, in sequence (don't parallelize — SerialInterface holds an exclusive port lock):**
- `device_info(port=<p>)` → `my_node_num`, `long_name`, `short_name`, `firmware_version`, `hw_model`, `region`, `num_nodes`, `primary_channel`
- `list_nodes(port=<p>)` → peer count, which peers have `publicKey`, SNR/RSSI distribution
- `get_config(section="lora", port=<p>)` → region, preset, channel_num, tx_power, hop_limit
- If anything looks off (can't connect, `num_nodes` wrong, missing `firmware_version`), open a short firmware-log window: `serial_open(port=<p>, env=<inferred>)`, wait 3 seconds, `serial_read(session_id, max_lines=100)`, `serial_close(session_id)`. Infer env from VID (0x239a → `rak4631`, 0x303a/0x10c4 → `heltec-v3`) unless an `MESHTASTIC_MCP_ENV_<ROLE>` env var overrides it.
4. **Render per-device report** as a compact block:
```text
[nrf52 @ /dev/cu.usbmodem1101] fw=2.7.23.bce2825, hw=RAK4631
owner : Meshtastic 40eb / 40eb
region/band : US, channel 88, LONG_FAST
tx_power : 30 dBm, hop_limit=3
peers : 1 (esp32s3 0x433c2428, pubkey ✓, SNR 6.0 / RSSI -24 dBm)
primary ch : McpTest
firmware : no panics in last 3s
```
Flag abnormalities inline with `⚠︎ <short reason>` — missing pubkey on a known peer, region UNSET, mismatched channel name, etc.
5. **Cross-device correlation** (when >1 device selected):
- Do both see each other in `nodesByNum`?
- Do `region`, `channel_num`, `modem_preset` match across devices?
- Do the primary channel names match? (Different name → different PSK → no decode.)
6. **Suggest next steps only for recognizable failure modes**, never speculatively:
- Stale PKI one-way → "`/mcp-test tests/mesh/test_direct_with_ack.py` — the test's retry+nodeinfo-ping heals this."
- Region mismatch → "re-bake one side via `./mcp-server/run-tests.sh --force-bake`."
- Device unreachable → refer operator to the touch_1200bps + CP2102-wedged-driver notes in `run-tests.sh`.
## Hard constraints
- **Read-only.** No `set_config`, no `reboot`, no `factory_reset`, no `flash`. If the operator wants mutation, they'll escalate explicitly.
- **Open/query/close per device.** Never hold multiple SerialInterfaces to the same port. The port lock is exclusive.
- **Don't infer env beyond the VID map** — if the operator has an unusual board, ask them which env to use rather than guessing.
.github/prompts/mcp-repro.prompt.md
+67
View File
@@ -0,0 +1,67 @@
---
mode: agent
description: Re-run a specific test N times to triage flakes; diff firmware logs between passes and failures (Copilot equivalent of the Claude Code /repro slash command)
---
# `/mcp-repro` — flakiness triage for one test
Equivalent of `.claude/commands/repro.md`. Use when the operator says "that one test is flaky — dig in", "repro the direct_with_ack failure", "why does X sometimes fail?".
## What to do
1. **Parse the operator's input** into two pieces:
- **Test identifier** — either a pytest node id (has `::` or starts with `tests/`) or a `-k`-style filter (plain substring like `direct_with_ack`).
- **Count** — integer, default `5`, cap at `20`. If the operator asks for 50, negotiate down and explain (airtime + USB wear).
2. **Sanity-check the hub** via the `list_devices` MCP tool. If the test name references `nrf52` or `esp32s3` and the matching VID isn't present, stop and report — re-running won't help.
3. **Loop** N times. Each iteration:
```bash
./mcp-server/run-tests.sh <test-id> --tb=short -p no:cacheprovider
```
`-p no:cacheprovider` keeps pytest from caching anything between iterations. Capture: exit code, duration, and (on failure) the `Meshtastic debug` firmware-log section from `mcp-server/tests/report.html`.
4. **Tally** results as you go:
```text
attempt 1: PASS (42s)
attempt 2: FAIL (128s) ← fw log captured
attempt 3: PASS (39s)
attempt 4: FAIL (121s)
attempt 5: PASS (41s)
--------------------------------------------------
pass rate: 3/5 (60%) | mean duration: 74s
```
5. **On mixed outcomes, diff the firmware logs** between one representative pass and one representative fail. Focus on:
- Error-level lines present only in failures (`PKI_UNKNOWN_PUBKEY`, `Alloc an err=`, `Skip send`, `No suitable channel`, `NAK`)
- Timing around the assertion point (broadcast sent? ACK received? retry fired?)
- Device-state fields that changed between attempts
Surface the top 3 differences as a compact "passes when / fails when" table with uptime timestamps. Don't dump full logs.
6. **Classify** the flake into one of:
- **LoRa airtime collision** — pass rate improves with fewer concurrent transmitters. Suggest a `time.sleep` gap or retry bump in the test body.
- **PKI key staleness** — first attempt fails, subsequent ones pass; existing retry-loop pattern in `test_direct_with_ack.py` is the fix.
- **NodeInfo cooldown** — `Skip send NodeInfo since we sent it <600s ago` in fail-only logs; needs a `broadcast_nodeinfo_ping()` warmup.
- **Hardware-specific** — one direction consistently fails, firmware versions differ, CP2102 driver wedged, etc.
- **Unknown** — say so. Don't invent a root cause.
7. **Report back** with:
- Pass rate + mean duration.
- Classification + the specific log evidence for it.
- A concrete next step (tighter assertion, more retries, open `/mcp-diagnose`, file a bug, nothing).
## Examples
- `tests/mesh/test_direct_with_ack.py::test_direct_with_ack_roundtrip[esp32s3->nrf52] 10` — 10 runs of that parametrized case.
- `broadcast_delivers` — no `::`, no `tests/`; treat as `-k broadcast_delivers`; runs every match 5 times.
- `tests/telemetry/test_device_telemetry_broadcast.py 3` — shorter count for a slow test.
## Notes
- If the FIRST attempt fails and the rest pass, that's a state-leak signature — suggest starting from `--force-bake` or a clean device state rather than chasing the first-failure firmware logs.
- If ALL N fail, this isn't a flake — it's a regression. Say so, stop iterating, escalate to `/mcp-test` for full-suite context.
- Don't rebuild firmware during triage. Flakes that only reproduce under different firmware belong in a separate session with a plan.
.github/prompts/mcp-test.prompt.md
+51
View File
@@ -0,0 +1,51 @@
---
mode: agent
description: Run the mcp-server test suite and interpret results (Copilot equivalent of the Claude Code /test slash command)
---
# `/mcp-test` — mcp-server test runner with interpretation
Equivalent of the Claude Code `/test` slash command in `.claude/commands/test.md`. Use this when the operator asks you to "run the tests", "check the mcp test suite", "run the mesh tests", etc.
## What to do
1. **Invoke the wrapper** from the firmware repo root:
```bash
./mcp-server/run-tests.sh [pytest-args]
```
If the operator specified a subset (e.g. "just the mesh tests"), pass it through as `tests/mesh` or a pytest `-k filter`. If they said nothing, use the wrapper's defaults (full suite with pytest-html report).
The wrapper auto-detects connected Meshtastic devices, maps each to its PlatformIO env, exports the required env vars, and invokes pytest. Zero pre-flight config needed from the operator.
2. **Read the pre-flight header** (first few lines of wrapper output). The `detected hub :` line lists role → port → env mappings. If it reads `(none)`, the wrapper narrowed to `tests/unit` only — call that out explicitly so the operator knows hardware tiers were skipped.
3. **On pass**: one-line summary like `N passed, M skipped in <duration>`. Don't enumerate test names. DO mention any non-placeholder SKIPs (things like "role not present on hub") because they indicate missing hardware or setup issues.
4. **On failure**: open `mcp-server/tests/report.html` (pytest-html output, self-contained) and extract the `Meshtastic debug` section for each failed test. That section includes a firmware log stream (last 200 lines) and device state dump. For each failure, summarise:
- test name
- one-line assertion message
- the specific firmware log lines that explain why (look for `PKI_UNKNOWN_PUBKEY`, `Skip send NodeInfo`, `Error=`, `Guru Meditation`, `assertion failed`, `No suitable channel`)
5. **Classify each failure** as one of:
- **Transient flake** — LoRa collision, first-attempt NAK with self-heal pattern, timing-sensitive assertion. Suggest `/mcp-repro <test-id>` to confirm.
- **Environmental** — device unreachable, port busy, CP2102 driver wedged on macOS. Suggest specific recovery (USB replug, `touch_1200bps`, `git status userPrefs.jsonc`).
- **Regression** — same assertion fails repeatedly on re-runs, firmware log shows novel errors. Identify the firmware module likely responsible.
6. **Do NOT run destructive recovery automatically**. If a failure looks like it needs a reflash, factory*reset, or replug — \_describe the steps* and let the operator decide. Never burn airtime or flash cycles without approval.
## Arguments convention
Operators generally invoke this prompt either with no arguments (full suite) or with a specific subset. Examples:
- `tests/mesh` — one tier
- `tests/mesh/test_direct_with_ack.py::test_direct_with_ack_roundtrip` — one test
- `--force-bake` — reflash devices first
- `-k telemetry` — name-filter
## Side-effects to confirm in your summary
- `userPrefs.jsonc` should be clean after a successful run. The session fixture in `mcp-server/tests/conftest.py` (`_session_userprefs`) snapshots and restores. Check `git status --porcelain userPrefs.jsonc` and report if it's non-empty.
- `mcp-server/tests/report.html` and `junit.xml` regenerate on every run.
- The wrapper prints a warning if a `.mcp-session-bak` sidecar was left over from a crashed prior session and auto-restores from it — mention that if it happened.