velocity.report: Skill for Claude Code

Skill: research-briefing

Discover what was published or newly surfaced in the week on roadside LiDAR perception and on the safety and behaviour literature the LiDAR analytics cite, evaluate each source against what the repository already holds, and commit one dated briefing on its own branch with a PR. The method is the one that found and judged the October 2026 sources: the Waymo sober-baseline and fatal-rate papers, Valgo's hosted crash baselines, the AAA Foundation attitudes and aggressive-driving reports, and FMCSA's handheld headway programme.

Usage

/research-briefing                              # the last 7 days, ending today
/research-briefing --since 2026-10-03           # a longer window
/research-briefing path/to/briefing.md          # also assess a supplied third-party briefing

The Monday routine "Weekly research briefing" invokes this skill in a fresh session with no arguments.

Procedure

1. Read the house rules first

AGENTS.md, TENETS.md, .github/STYLE.md and docs/platform/operations/documentation-standards.md. The output is one briefing and, where its own verdict says "adopted", a definitions paragraph, an evaluation-source row, a reference or an open question in the plan it affects. Nothing larger changes: anything larger is a proposal inside the briefing.

2. Window

Primary work published or first listed in the last 7 days (or since --since), plus older work newly surfaced this week: an agency project page updated, a preprint accepted, a dataset or API newly announced. Say which of the two a source is. If nothing new was published, say so in one sentence and still record what surfaced.

3. Scope

Two bodies of literature the project depends on:

1. Roadside and infrastructure LiDAR perception: detection, clustering, tracking, extent and heading estimation, occlusion and fragmentation, multi-object identity, background modelling, calibration and ground-truth methods, and trajectory datasets (highD, inD, rounD, openDD, Waymo Open Motion, and any new one).
2. The safety and behaviour literature the analytics cite: headway, following distance, clearance and net time gap, passage headway, post-encroachment time, time to collision, stop compliance, speed percentiles and p85 practice, speeding exposure and free-flow opportunity, crash involvement rates and exposure denominators (vehicle-miles, FARS, police-reported), sober and status-quo human baselines for automated-driving comparison, pedestrian impact-speed harm curves, speed-change crash models, and public attitudes to speeding and aggressive driving.