Queries project documentation to answer questions about architecture and component usage.
Install
mkdir -p .claude/skills/qa-charon349 && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/17371" && unzip -o skill.zip -d .claude/skills/qa-charon349 && rm skill.zipInstalls to .claude/skills/qa-charon349
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Answer questions about the project by reading documentation sources in priority order. Use when the user asks "how does X work", "where is X", "what does X do", or any project-related question.Key capabilities
- →Answer questions about project structure
- →Answer questions about high-level project description and usage
- →Answer questions about development workflow and testing
- →Answer questions about detailed Sphinx documentation pages
- →Answer questions about specific topics like data generation or training
How it works
The skill reads documentation sources in a prioritized order, starting with general project overviews and then diving into topic-specific files to answer user questions.
Inputs & outputs
When to use qa
- →Find where a feature is implemented
- →Explain how a component works
- →Get setup instructions
About this skill
Project Q&A
Answer the user's question about this project. Use the general lookup order first, then the topic-specific maps below for deeper dives.
General lookup order
Read these in order:
.claude/docs/project_structure.md— directory layout and component overview.README.md— high-level project description, setup, and usage.DEVELOPING.md— development workflow, testing, and contribution guide.docs/source/— detailed Sphinx documentation pages.
Topic: Data generation / collection
- slurm/data_collection/collect_data.py — SLURM orchestrator launching parallel CARLA data collection jobs.
- slurm/data_collection/delete_failed_routes.py — cleans up failed route directories.
- 3rd_party/leaderboard_autopilot/leaderboard/leaderboard_evaluator_local.py — runs CARLA scenarios per route.
- lead/main.py — adapts LEAD to the CARLA leaderboard agent protocol.
- lead/expert/expert.py — main expert agent driving logic.
- lead/expert/expert_base.py — shared base class for expert variants.
- lead/expert/expert_data.py — saves frames and labels during collection.
- lead/expert/config_expert.py — expert configuration dataclass.
- lead/expert/privileged_route_planner.py — privileged route/lane planning.
Topic: Data preparation
- lead/data_loader/carla_dataset.py — CARLA Leaderboard 2.0 PyTorch dataset.
- lead/data_loader/carla_dataset_utils.py — CARLA loading and preprocessing helpers.
- lead/data_loader/training_cache.py — compressed on-disk training cache.
- lead/data_loader/navsim_dataset.py — NavSim PyTorch dataset.
- lead/data_loader/waymo_e2e_dataset.py — Waymo Open Dataset end-to-end dataset.
- lead/data_buckets/abstract_bucket_collection.py — ABC for bucket collections (serialization, loading).
- lead/data_buckets/bucket.py — single bucket holding paths and metadata.
- lead/data_buckets/route_filtering.py — route validity and deduplication.
- scripts/build_cache.py — entry point to build the training cache.
- scripts/build_buckets_pretrain.py — generate pretraining bucket collections.
Topic: Training
- lead/training/train.py — main DDP-aware training entry point (Trainer class).
- lead/training/config_training.py — training hyperparameters dataclass.
- lead/training/training_utils.py — seeding, checkpointing, data loader construction.
- lead/training/mixed_training_utils.py — multi-dataset sample scheduling.
- lead/training/logger.py — WandB metric and media logging.
- scripts/pretrain_ddp.sh — DDP pretraining launcher.
- scripts/posttrain_ddp.sh — DDP posttraining launcher.
- slurm/experiments/ — per-experiment SLURM scripts (see
/prepare-slurm-scripts).
Topic: Model architecture (TransFuser v6)
- lead/tfv6/tfv6.py — top-level model orchestrating backbone and decoder heads, loss aggregation.
- lead/tfv6/transfuser_backbone.py — image + LiDAR fusion backbone.
- lead/tfv6/transfuser_utils.py — backbone utility layers and functions.
- lead/tfv6/planning_decoder.py — waypoint, path and target speed planning head.
- lead/tfv6/bev_decoder.py — BEV semantic segmentation head.
- lead/tfv6/perspective_decoder.py — perspective view output head.
- lead/tfv6/center_net_decoder.py — CenterNet 2D bounding box detector.
- lead/tfv6/radar_detector.py — 2D radar object detector.
Topic: Evaluation (bench2drive, longest6, town13)
- slurm/evaluation/evaluate.py — core orchestrator: submit, monitor, retry SLURM eval jobs.
- slurm/evaluation/evaluate_scripts_generator.py — generates per-route shell scripts.
- slurm/evaluation/evaluate_utils.py — JSON parsing, status checks.
- slurm/evaluation/merge_route_json.py — merges per-route results into one summary.
- slurm/evaluation/evaluate_wandb_logger.py — logs aggregated eval results to WandB.
- lead/inference/closed_loop_inference.py — CARLA closed-loop evaluation with PID control.
- lead/inference/open_loop_inference.py — open-loop evaluation (NavSim / Bench2Drive).
- lead/inference/sensor_agent.py — sensor-based agent wrapper for the CARLA leaderboard protocol.
- lead/inference/infraction_recorder.py — records and serializes driving infractions.
- lead/inference/config_closed_loop.py — closed-loop eval config.
Rules
- Read in order. Start with general docs, then dive into topic-specific files.
- Be specific. Point the user to the exact file, section, or line.
- Don't guess. If none of the sources answer the question, say so and suggest where else to look.
When not to use it
- →When the question is not about the project
- →When the question requires guessing information not present in sources
Limitations
- →The skill only uses provided documentation sources
- →The skill will not guess if information is not found in sources
How it compares
This skill provides specific, source-grounded answers by following a defined lookup order, unlike a general search or an unguided response.
Compared to similar skills
qa side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| qa (this skill) | 0 | 2mo | No flags | Beginner |
| citation-management | 23 | 7mo | Review | Beginner |
| microsoft-docs | 10 | 4mo | Review | Beginner |
| peer-review | 11 | 7mo | Review | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
citation-management
davila7
Comprehensive citation management for academic research. Search Google Scholar and PubMed for papers, extract accurate metadata, validate citations, and generate properly formatted BibTeX entries. This skill should be used when you need to find papers, verify citation information, convert DOIs to BibTeX, or ensure reference accuracy in scientific writing.
microsoft-docs
github
Query official Microsoft documentation to understand concepts, find tutorials, and learn how services work. Use for Azure, .NET, Microsoft 365, Windows, Power Platform, and all Microsoft technologies. Get accurate, current information from learn.microsoft.com and other official Microsoft websites—architecture overviews, quickstarts, configuration guides, limits, and best practices.
peer-review
davila7
Systematic peer review toolkit. Evaluate methodology, statistics, design, reproducibility, ethics, figure integrity, reporting standards, for manuscript and grant review across disciplines.
notion-research-documentation
openai
Research across Notion and synthesize into structured documentation; use when gathering info from multiple Notion sources to produce briefs, comparisons, or reports with citations.
context7-auto-research
davila7
Automatically fetch latest library/framework documentation for Claude Code via Context7 API
fact-check
leonardomso
Verify technical accuracy of JavaScript concept pages by checking code examples, MDN/ECMAScript compliance, and external resources to prevent misinformation