Menu
Use when drafting, executing, or reviewing Codex super-prompts for this repository.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when ingesting, validating, curating, or querying production sensor data as a versioned dataset-of-record, or wiring the dataset-ingest-curate workflow.
Use when generating adversarial scenarios via RL, ranking mined failures of a policy-under-test, or wiring the adversarial-scenario-hardening workflow.
Use when navigating the Cosmos3 integration in NPA or locating upstream Cosmos3 framework files, defaults, scripts, configs, recipes, and docs.
Use when Cosmos3 setup, fetch, inference, CUDA, uv, Docker, Hugging Face, GitHub, NGC, or checkpoint staging fails in NPA or in an upstream Cosmos framework checkout.
Use when setting up Cosmos3 access through NPA, checking source or Hugging Face reachability, staging the public Cosmos3 framework and checkpoint cache, or deciding which NPA workflow to use before inference.
Use for Claude Code reviews that need robotics, simulation, GPU-routing, sim-to-real, or BDD100K pipeline domain context.
| name | super-prompt-patterns |
| description | Use when drafting, executing, or reviewing Codex super-prompts for this repository. |
Use RELAXED_DIRTY_TREE_MODE: halt only when dirty files overlap with the run's target paths. Pre-existing dirty files in unrelated paths never halt the run.
Do not add time, cost, or job-count limits unless the operator explicitly specifies them. Strip per-phase time budgets, wall-clock caps, and GPU spend limits from prompts.
Standard phase flow:
Log novel issues to /tmp/<run-id>/novel-issues.md. Use log/skip/continue for non-blocking issues and halt only for true blockers.
Acquire /tmp/npa-commit-lock/<scope> during commit phases for parallel runs that share files.
If the npa CLI is broken during a run, document the breakage as a structured prompt, run Codex to fix it, then use the fixed CLI.
Use separate agents by phase: Build with Codex, verify with Codex in read-only mode, review with Claude Code for architecture, then cleanup with Codex in a scoped pass.
Trigger a Claude Code review after 3 or more commits land. The CC agent must differ from the building agent.
Root AGENTS.md is a lightweight index; details belong in skill files.
