mit einem Klick
codex_skills
codex_skills enthält 52 gesammelte Skills von Idate96, mit Repository-Berufsabdeckung und Skill-Detailseiten auf SkillsMP.
Skills in diesem Repository
Current Moleworks Terra trenching runbook for full autonomous Beam6-style trench execution in simulation or on the robot. Use when investigating or running the two-stage flange/bottom trench flow, generate_trench_sequence_plans.py, beam6_sequence_stage.launch.py, BASE_CONTROL target registration, mesh_to_excavation_grid_map.py, workspace planner trench-axis metadata, Terra behavior-tree activation, Newton or Isaac/Terra simulation bringup, robot bringup, and 400 mm tool handoff.
Read recent Google Chat context, draft or send a reply in the correct DM or space, download collaborator attachments such as timesheets or PDFs, and handle simple meeting coordination by creating or updating a Google Calendar invite and posting the Meet link back in Chat. Use when Lorenzo asks to read a collaborator's recent messages, understand chat context before replying, send a Google Chat reply through the Chat API, pull a PDF or timesheet out of Chat, or create a meeting from a chat exchange.
Replay split DIG bags in the `moleworks_ros` container with bag TF, live self-filter, live elevation mapping, live excavation mapping, and Foxglove. Use when reviewing DIG episodes from `sensors/`, `state/`, `commands/`, `lidar/`, and optional `elevation_map/` bags.
Finalize RSL student grading and offboarding. Use when Lorenzo asks to find a student's grading sheet, extract or submit a grade, update the RSL student-project tracker like the onboarding workflow, request eDoz grade entry from admin staff, mark offboarding fields such as completed/report/grading/source/access-revoked only with evidence, or send a short Google Chat status reply after the handoff.
Validate the Newton + ROS Nav2 driving stack in a clean tmux session after bringup. Use when the user wants a repeatable navigation check in Newton sim, including health checks for the bridge/model/drive path and the lateral-shift golden test.
Start or restart the Moleworks ROS2 stack using the Newton simulator in the default moleworks_ros runtime shell, assuming the current shell is already inside the target container unless the user says otherwise. Use when you need a clean tmux layout for Newton bridge, robot/TF/RViz, perception (elevation + excavation mapping), optional Foxglove bridge, an isolated bridge-only validation stack on a specific ROS domain, or Terra failure capture and resume from saved checkpoints in Newton simulation, all with use_sim_time:=true.
Start/restart Mole OCS2 arm controller experiments with real actuation commands in tmux, use configure-time auto-bootstrap (hold+reset+first-policy), activate safely, send cylindrical (r, theta, z, pitch) end-effector goals (Foxglove or CLI), record rosbag, and run staged validation while preventing competing /mole/actuator_commands publishers (e.g. dig_controller).
Use the @steipete/oracle CLI to bundle a prompt plus the right files and get a second-model review (API or browser) for debugging, refactors, design checks, or cross-validation.
Submit, monitor, sync, and ledger Moleworks IsaacLab RL runs in `moleworks_ext`. Use when preparing a smoke test, launching Euler jobs, debugging failed Slurm runs, syncing experiments, comparing run configs, or updating `docs/EXPERIMENTS_ONGOING.md` and `docs/EXPERIMENTS_RUN.md`.
Benchmark and analyze Moleworks Newton RL checkpoints. Use when benchmarking analytic cabin or shared-turn checkpoints, collecting or replaying terrain banks, verifying a saved bank, comparing fresh vs carved terrain, or summarizing benchmark JSON outputs.
Submit, monitor, sync, and ledger Moleworks Newton RL runs on Euler or Brev. Use when preparing a smoke test, launching shared-turn jobs, checking `squeue`/`sacct`/Slurm logs, syncing one run, or updating `docs/experiments`.
Entry point for Moleworks Newton RL work. Use when working on Newton training, shared-turn experiments, local smoke tests, cluster launches, benchmark interpretation, or experiment ledgers. Routes to narrower Newton skills for cluster ops, benchmarking, ROS parity, and long-horizon orchestration.
Set up the standard Moleworks ROS 2 tmux session for on-machine work with the base stack windows (low_level, perception, estimator, foxglove), and optionally launch a dig controller when the user asks for base stack plus dig. By default also unlock hydraulics and set engine RPM to 1600 unless the user asks otherwise.
ROS2 debugging with proper TF buffer timeouts. Use when checking transforms, topics, nodes, services, or tmux-managed ROS2 systems.
Launch, validate, compare, and monitor Terra / terra-baselines PPO training jobs on the Euler cluster. Use when Codex is asked to start Terra policy training, compare W&B hyperparameters, run single-excavator or mixed-agent `train_mixed.py` jobs, prepare Slurm scripts, check `squeue`/`sacct` logs, or document/default Terra cluster training parameters.
Execute packaged Moleworks Terra flat-foundation plans in Newton simulation, especially flat_foundation_depth_0p5 end-to-end with Nav2 navigation, async checkpoints, workspace-planner target recomputation, post-dump stall measurement, checkpoint resume, and failure-state capture. Use after newton-sim-ros-startup has established the ROS/Newton runtime.
Run, validate, debug, and document Terra RL training in the terra and terra-baselines repositories. Use for Euler Slurm jobs, W&B run comparison, PPO/default solo-excavator configs, JAX/CUDA runtime preflight, first-update smoke tests, experiment ledgers, and diagnosis of slow compilation or GPU training failures.
Benchmark Moleworks workspace planner heuristics using the standardized HTML bundle generator. Use when regenerating workspace planner benchmark bundles, leaderboards, manifest/progress CSVs, trench and fan visualizations, or comparing free-lane pointwise_score and max_strip_volume planners.
Debug Moleworks workspace planner Newton/ROS runs with per-action planner GridMaps, predicted-vs-executed scoop analysis, Terra checkpoint replay of failed or high-discrepancy scoops, and live-run inspection artifacts.
Start Moleworks dig controllers (dig_3d, dig_newton, dig_ee, etc.) in a split tmux window: one pane runs the controller launch, the other pane runs lifecycle helpers and shows the `ros2 action send_goal` command. Use when the base stack is already running and you want to add or restart only a dig controller.
Bring up and test the Moleworks Terra ROS2 + IsaacLab stack in the correct Docker containers and tmux windows; use when starting/restarting isaac-terra + ros-terra, setting ROS_DOMAIN_ID and optional DDS overrides, and validating /mole/state, /clock, TF, and basic node health.
Start or restart the Moleworks IsaacLab simulation and ROS2 stack in a clean, repeatable tmux layout (sim, robot, perception). Use when you need a fresh sim + TF/robot visualization + perception (including excavation mapping) in RViz.
Record or generate dense 1 cm Open3D/Foxglove-style camera-colored 3D maps from Moleworks ROS 2 LiDAR and camera streams. Use when asked for Open3D mapping, colored maps, dense site maps, survey clouds, camera-colored LiDAR maps, or syncing generated map artifacts.
Start fast split rosbag recording for Mole dig/newton runs using the canonical `rosbag_record.launch.py` workflow. Use when you want separate bags for sensors, state, commands, lidar, camera (compressed image topics), elevation_map, and Dig3D special observations during digging experiments.
Scan Lorenzo's recent Google Chat activity at the start of the day and turn it into a concise todo list. Use when Lorenzo asks for a morning chat scan, chat triage, "what do I need to do from chat", "scan my chat and make a todo list", or similar. Reads recent collaborator DMs from the collaborator note and recent active smaller spaces via the Google Chat API, then extracts action items, reply obligations, and deadlines.
Stage and upload files to Kleinkram with mission creation and verification, or manage project/user access for Kleinkram datasets. Use when asked to upload datasets, run bags, or sysid artifacts to Kleinkram, or when asked to add users, inspect project access, or update Moleworks project permissions.
Sign a local PDF contract or form by placing an existing handwritten signature image onto one or more pages, then visually verify the result. Use when Lorenzo asks to sign a PDF in Downloads, adjust signature placement, or reuse an existing local signature image for a fast non-cryptographic signature workflow.
Connect to `lorenzo@starship` over Tailscale and verify that this machine is on the `lorenzoterenzi96@gmail.com` tailnet. Use when the user asks to SSH into `starship`, `starship one`, or the previous `starship-1` name, confirm access from this workstation, or open an interactive shell after Tailscale browser approval.
Orchestrate sub-agents for long-horizon Moleworks code, RL, and ROS tasks. Use when the task needs many code/doc searches, Slurm or W&B inspection, benchmark comparisons, or ROS process inspection without polluting the parent context.
Validate or replay Moleworks Newton observations through `moleworks_ros`. Use when bringing up the ROS parity stack, checking `/clock` and TF, debugging controller-facing terrain topics, replaying Dig3D-style comparisons, or cleaning up stale ROS/Newton processes after parity work.
Benchmark and analyze Moleworks IsaacLab RL checkpoints. Use when playing a policy locally, benchmarking excavation checkpoints, ranking multiple checkpoints fairly, plotting TensorBoard progression, or comparing synced run artifacts before recommending a checkpoint.
Entry point for Moleworks IsaacLab RL work in `moleworks_ext`. Use when working on local smoke tests, Euler submits, run sync, policy playback, benchmark interpretation, or experiment tracking. Routes to narrower IsaacLab skills for cluster ops and benchmarking.
Launch and debug Moleworks IsaacLab commands inside the moleworks_ext Docker container with persistent tmux windows and captured output. Use when starting containerized training/play/eval scripts and when validating command startup.
Restack, verify, and merge stacked Moleworks `moleworks_ros` PRs, including worktree setup, rebases onto `main`, local Docker CI on this machine, GitHub check monitoring, and final merge/cleanup. Use when a user asks to update a stack after `main` changed, split a monolithic PR across follow-up PRs, fix CI on several related branches, or merge the remaining open Moleworks ROS PRs.
Interact with local Chrome browser session (only on explicit user approval after being asked to inspect, debug, or interact with a page open in Chrome)
Use Google Chat in the local Chrome session through chrome-cdp. Use when asked to open Google Chat, verify the active Google account, switch to an existing Google account manually if needed, find a DM or space, and send or inspect messages in the already logged-in browser.
Onboard a new RSL student project across the student spreadsheet, shared Google Drive project folders, and follow-up access/admin requests. Use when Codex must add a new student entry, create the project folder from the template, place the grading sheet correctly, or follow the ETH/RSL student-start checklist.
Write and review short-lived research code with an aggressively simple, skimmable style. Use when the user asks for simpler code, fewer states, narrower APIs, discriminated unions, exhaustive handling, assertions instead of defensive code, or pre-PR cleanup that removes non-essential changes and complexity.
Recollect bounded-joint open-loop current LUT data on the robot machine using one bag per amplitude, asymmetric start poses, dense low-speed knee filling, and immediate per-point analysis. Use when recollecting or repairing `J_TELE` / `J_EE_PITCH` LUT points or when `mole_sysid_tune_lut` rejects edge amplitudes.
Edit and restructure PowerPoint presentations (.pptx) with an iterative render-and-inspect loop. Use when you must preserve existing media (images, GIFs, embedded videos), update slide text/layout, reorder slides, or export slide previews (PDF/PNGs) for visual review.