{"id":"7627f66c-87fa-4ce8-94a5-01960337c6d3","slug":"i4h-workflow-dataset-replay","kind":"skill","name":"i4h-workflow-dataset-replay","summary":"Replay a workflow HDF5 episode through its original Scene. Use for visual trajectory and recording verification; do not use for policy evaluation or LeRobot data.","owner":"NVIDIA","repoUrl":"https://github.com/NVIDIA/skills/tree/HEAD/skills/i4h-workflow-dataset-replay","license":"Apache-2.0","licenseTier":"permissive","trust":"official","stars":3586,"match":null,"fewStars":false,"format":"skill","createdAt":null,"addedAt":"2026-10-14T05:44:39.784Z","logoUrl":"https://avatars.githubusercontent.com/u/1728152?v=4","linkOnly":false,"collection":{"name":"skills","repo":"nvidia/skills"},"skillPath":"skills/i4h-workflow-dataset-replay","categories":["developer-tools","data-analytics"],"tags":["agentic-ai","ai-agents","claude-code","codex","coding-agents","cuda","developer-tools","llm"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-14T05:44:39.783Z","sample":false,"url":"https://conduits.dev/tools/i4h-workflow-dataset-replay","language":"Python","transports":[],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# Replay a Workflow Recording\n\n## Purpose\n\nReplay the exact recorded action sequence through the matching workflow and inspect it visibly.\n\n## Instructions\n\n1. Run the checkout resolver and select the exact HDF5.\n2. Read the original workflow and requested episode.\n3. Inspect action-contract compatibility.\n4. Run the replay visibly in the foreground through completion.\n\n## Resolve the recording\n\n```bash\nexport I4H_WORKFLOWS_REPO_URL=\"${I4H_WORKFLOWS_REPO_URL:-https://github.com/isaac-for-healthcare/i4h-workflows}\"\nI4H_REPO_DIR_NAME=\"${I4H_WORKFLOWS_REPO_URL%/}\"\nI4H_REPO_DIR_NAME=\"${I4H_REPO_DIR_NAME##*/}\"\nI4H_REPO_DIR_NAME=\"${I4H_REPO_DIR_NAME##*:}\"\nI4H_REPO_DIR_NAME=\"${I4H_REPO_DIR_NAME%.git}\"\n[ -n \"$I4H_REPO_DIR_NAME\" ] || { echo \"Cannot derive a checkout name from I4H_WORKFLOWS_REPO_URL\" >&2; exit 2; }\nROOT=\"${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}\"\nif [ ! -d \"$ROOT/workflows/i4h_workflows\" ]; then\n  ROOT=\"${I4H_WORKFLOWS:-$HOME/$I4H_REPO_DIR_NAME}\"\n  [ -d \"$ROOT/workflows/i4h_workflows\" ] || git clone \"$I4H_WORKFLOWS_REPO_URL\" \"$ROOT\"\nfi\nexport I4H_WORKFLOWS=\"$ROOT\"\ncd \"$ROOT\"\nfind runs -name '*.hdf5' -type f -printf '%T@ %p\\n' | sort -nr | head\n```\n\nTreat the resolver above as part of the skill contract: a hosted copy may run outside the base repository, so never assume the current checkout contains `workflows/i4h_workflows`. `I4H_WORKFLOWS_REPO_URL` selects the clone source. When `I4H_WORKFLOWS` is unset, derive the fallback directory from that URL; set `I4H_WORKFLOWS` only to reuse or choose a specific destination. Never replace an existing checkout.\n\nUse an explicit or current-chain recording first. Otherwise select the newest plausible HDF5 and state that choice. Never substitute an older file after a failed recording.\n\nInterpret natural ordinals as zero-based indices: first is `0`, second is `1`.\n\n## Inspect before launch\n\n```bash\nHDF5_PATH=/absolute/path/to/recording.hdf5\nuv run --project tools/dataset i4h-dataset inspect \"$HDF5_PATH\" --segments\n```\n\nResolve the original workflow from recording metadata and conversation context. Confirm the requested episode exists and its action width matches the workflow's replay Scene contract.\n\n## Replay\n\n```bash\n./run.sh <workflow> --replay \"$HDF5_PATH\" --episode <zero-based-index>\n```\n\nKeep the visible simulator and command in the foreground. Poll yielded execution until `run.sh` exits; do not detach or return while replay is still running.\n\n## Verify\n\nObserve the complete trajectory, relevant objects, camera views, segment boundaries, and final status. If motion diverges, report the first mismatching segment or action-contract error. Do not change workflow, episode, or recording silently.\n\n## Troubleshooting\n\nOn launch failure, verify workflow metadata, episode existence, and action width. On divergence, report the first mismatching segment.\n\n## Prerequisites\n\nRequire the original workflow assets and a readable episode whose action width matches replay mode.\n\n## Limitations\n\nReplay verifies stored actions in one Scene; it does not evaluate a learned policy or guarantee transfer to another Scene.\n\n## Examples\n\n- `Replay the second episode.` → select the current recording, map “second” to index `1`, and observe the visible replay.\n\n## Completion gate\n\nReport workflow, HDF5 path, episode index, frame/segment count, action width, exit/final status, and visible agreement or first mismatch."}