The drive phase gated on wall-clock time but asserted a sim-time
distance. The sim does not run at realtime (scans arrive well above the
configured rate), so on a loaded machine a wall-clock duration yields a
variable distance and could fail the assertion spuriously. Gate the
drive on sim time (with a wall-clock safety cap so a stalled /clock
can't hang), and compute scan rate from message header stamps. Forward
distance now lands at 0.600 m exactly.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Record why the sensor+SLAM smoke test can't run on GPU-less hosted
runners: software rendering starves the gz loop and the in-process
controller_manager can't spawn controllers in time.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

The sim and sim-test tasks are defined only in the sim feature, so
'pixi run sim' / 'pixi run sim-test' auto-select that environment (same
as 'pixi run rviz' for the dev env). Only ad-hoc 'pixi run -- <cmd>'
invocations still need -e sim, since those default to the robot env.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

A quick pre-commit setup (~1 s cached): whitespace/EOF/line-ending
fixers, YAML/TOML validation, merge-conflict and large-file guards,
shellcheck for the shell scripts, and ruff (pyflakes only, no style
reformatting) for the Python launch files. Slow build-dependent checks
stay in CI / 'pixi run test' / 'pixi run sim-test'.

Runs in its own minimal 'lint' pixi environment (no ROS) so it solves
and installs fast. Tasks: 'pixi run lint' and 'pixi run lint-install'.
design/ and docs/images/ are excluded so text hooks never rewrite the
ASCII STEP CAD files or binary assets. verify_sim.py marked executable
to satisfy the shebang check.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Track A of the dev-story plan. watchexec (dev feature) continuously
rsyncs to the Pi on save via 'pixi run sync-watch'. README now spells
out the colcon --symlink-install fast path: edits to launch/config/
Python go live with no rebuild, so the loop is sync-watch + re-launch,
not sync + rebuild. Replaces the stale 'pixi pack TBD' note with the
pixi-build-ros channel direction.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

design/WIRING.md: connection tables (power + data), a GitHub-rendered
Mermaid topology diagram, the 5V power analysis (servos run below their
7.4V rating; USB-C PD delivers far less than its rated wattage at 5V,
which is why the servo board needs the high-current Out1 port), a rough
5V budget, and the BNO085 I2C pinout. design/ASSEMBLY.md: per-part print
settings and step-by-step assembly grounded in the 235mm chassis / 50mm
standoff / power-bank-sandwich / ORP-grid design. Both linked from
design/README.md.

Hardware-specific unknowns (connector genders, servo stall current at
5V, exact screw lengths, tyre material) are flagged as Verify rather
than guessed.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Reframe the servo-voltage note: 5V operation is confirmed working, not a
performance worry. Keep only the factual caveat that datasheet torque
figures are quoted at the higher rated voltage.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Tyre is TPU (confirmed). Screws: most joints M3x12 (~6mm plate + ~6mm
captive-nut pocket), M3x10 for thinner stacks. All data links are plain
USB, so the connector-gender Verify flags are resolved. Remaining Verify
items are electrical only (barrel-cable 5V profile, servo stall current,
BNO085 VIN voltage).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>