AMDResearch

Run cleanroom build+test from build tree, install tree, and relocated install tree.

Create a structured decision workflow for design-heavy repositories. Use when the user needs to compare options with tradeoff analysis, assign a decision owner, and track consequences. Produces a DW-NNN document with options, pros, cons, and tradeoff analysis.

Turn a rough design discussion into a structured written brief. Use when the user has unstructured notes, chat logs, or conversation fragments that need to become a clear document with positions, open questions, and decision points.

Synchronize documentation indices and cross-references. Use after adding, renaming, moving, or deleting documentation files. Finds orphaned docs, stale links, and broken cross-references, then fixes them and reports changes.

Collect user feedback about the Agentic Meta Project and file it. Use when the user says "feedback", "report a bug", "file an issue", or wants to report friction, suggestions, or feature requests. Files as a GitHub issue if gh CLI is available, otherwise writes a local report.

Package a locally discovered lesson for forwarding to the Agentic Meta Project. Use after a session review identifies a finding that generalizes beyond this project -- a template gap, missing skill, or structural issue that would affect any project using the system.

Create or manage a bug-fix workflow packet. Use when the user reports a bug, says "fix this bug", or needs a structured workflow for diagnosing and resolving a defect. Emphasizes reproduction, root cause analysis, minimal fix, and regression testing.

Create or manage a feature workflow packet. Use when the user wants to add new user-visible functionality and needs a tracked workflow with acceptance criteria, scope boundaries, and a verification strategy.

Create a structured decision workflow for design-heavy repositories. Use when the user needs to compare options with tradeoff analysis, assign a decision owner, and track consequences. Produces a DW-NNN document with options, pros, cons, and tradeoff analysis.

Turn a rough design discussion into a structured written brief. Use when the user has unstructured notes, chat logs, or conversation fragments that need to become a clear document with positions, open questions, and decision points.

Synchronize documentation indices and cross-references. Use after adding, renaming, moving, or deleting documentation files. Finds orphaned docs, stale links, and broken cross-references, then fixes them and reports changes.

Collect user feedback about the Agentic Meta Project and file it. Use when the user says "feedback", "report a bug", "file an issue", or wants to report friction, suggestions, or feature requests. Files as a GitHub issue if gh CLI is available, otherwise writes a local report.

Package a locally discovered lesson for forwarding to the Agentic Meta Project. Use after a session review identifies a finding that generalizes beyond this project -- a template gap, missing skill, or structural issue that would affect any project using the system.

Create or manage a bug-fix workflow packet. Use when the user reports a bug, says "fix this bug", or needs a structured workflow for diagnosing and resolving a defect. Emphasizes reproduction, root cause analysis, minimal fix, and regression testing.

Create or manage a feature workflow packet. Use when the user wants to add new user-visible functionality and needs a tracked workflow with acceptance criteria, scope boundaries, and a verification strategy.

Initialize or refresh Project Memory for a repository. Use after amplifying a new repo, when the user says "init PM", "set up project memory", or when pm-index.md is missing or empty. Surveys the codebase, creates starter PM units, and populates the index.

Applies to Earth-system ML in AI4Science Studio—climate, weather, geospatial data, and model folders under earth_science/models/. Includes validated AMD/ROCm HPC patterns for earth science models.

Applies to material science and chemistry ML recipes under material_science/models/ in AI4Science Studio.

Runs the AMD AI agents bottleneck-analysis workflow on a HydraGNN or ORBIT-2 training job using TraceLens + Omnistat user-mode + paired analyst/verifier subagents, or drives the iterative perf-optimizer-loop (accept/revert on throughput or epoch time). Use when the user wants to analyze a perf run, analyze a 2-node HydraGNN run, find bottlenecks with TraceLens/Omnistat, run the AMD AI agents on a job, run the perf-optimizer-loop, tune ORBIT-2 throughput, or optimize HydraGNN training performance.

Applies when working in the AI4Science Studio repository. Describes domain layout, model slug rules, where recipes live, safety expectations, and AMD/ROCm HPC patterns (Apptainer and Docker) validated in practice.

Step-by-step instructions for how Cursor should run any model in AI4Science Studio — read model.yaml, check preflight, select container runtime, launch.

Applies to healthcare and life sciences ML recipes under healthcare/models/ in AI4Science Studio, with privacy and non-clinical-use guardrails.

How to answer user questions about what models are available, filter by domain/task/license, and compare models in AI4Science Studio.

Guides creation and editing of Hugging Face–centric recipes in AI4Science Studio—discovery, task alignment, scripting, documentation, and attribution.

Generates SLURM job scripts with omnihub-generate-job and runs them with sbatch. Use when the user wants to generate a job, run a job, use omnihub-generate-job, sbatch, set app-config, partition, or run a single OmniHub job.

Documents OmniHub repo layout, app configs, cluster and tools config, results layout, and when to use runner manual vs torchrun. Use when the user asks about conventions, app config, cluster config, tools config, results layout, or runner manual vs torchrun.

Creates and runs parameter sweeps with omnihub-sweep using templates, sweep-dir, and optional resume. Use when the user wants to run a sweep, use omnihub-sweep, create a parameter sweep, use a sweep template, set sweep-dir, or do a dry-run sweep.

Periodically checks the results of the current or specified OmniHub job: inspects results_dir and logs, alerts on errors, and reports success. Use when the user wants to monitor a job, check for errors in results or logs, see job status, or get a success/failure report for an OmniHub job.

Primus pretraining via OmniHub -- job generation, config overrides, MASTER_PORT, HF_HOME, GPU arch (FP8), and debugging multinode hangs. Use when the user works with applications/primus-pretrain/, pretrain_wrapper.py, Primus config, primus_turbo, or multinode Primus training.

Processes and indexes OmniHub job results with omnihub-process and omnihub-index to produce index.csv and per-job processed-data. Use when the user wants to post-process results, run omnihub-process, omnihub-index, build index.csv, or work with processed-data.

Profile GPU kernels when performance analysis or optimization is required. Use for AMD ROCm GPU metrics, bandwidth, cache hit rates, coalescing, or kernel timing.

Test local kerncap changes end-to-end by profiling an application, extracting a kernel, and validating the reproducer. Use when the user asks to test kerncap against any HIP or Triton workload, or wants to validate extraction on a real GPU application.

Validate GPU kernel correctness by comparing reference and optimized outputs. Use when verifying that an optimized or modified kernel matches a reference implementation.

Profile GPU kernels at source-line granularity with cycle-level timing and stall analysis. Use when identifying performance hotspots at the source code level or analyzing instruction-level metrics mapped to source lines.

Extract GPU kernel assembly and HIP source from HSA packet traces. Use when analyzing what code ran on the GPU, debugging kernel dispatch, or inspecting assembly and source mapping.