CtrlK
BlogDocsLog inGet started
Tessl Logo

running-openmed-ondevice

Run OpenMed models fully on-device with the MLX (Apple Silicon), CoreML (iOS/macOS), or ONNX/WebGPU (cross-platform/browser) backends, including convert-quantize-run workflows. Use when the user wants to deploy OpenMed at the edge, run NER/de-id on Apple Silicon, target iPhone/iPad/Mac, export to ONNX or WebGPU, quantize a clinical model to int8/4-bit, run with no network, or pick between MLX/CoreML/ONNX. Covers the mlx/coreml/onnx extras, the convert() functions and python -m convert CLIs, quantization, loading a local MLX artifact through analyze_text, OpenMedMLXLanguageModel/generate_text, and the on-device-only PHI guarantee (nothing leaves the host).

76

Quality

94%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

88%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A code-dense, highly actionable skill body with a clear sequenced workflow and explicit validation/eval gates appropriate for clinical de-id. Minor conciseness gains are available in the intro, and progressive disclosure is solid but relies entirely on inline content with no reference files.

Suggestions

Trim the intro paragraph's backend-role restatement since the 'Pick a backend' table already conveys it, saving tokens.

Consider moving the per-backend gotchas or the LLM-generation detail into a reference file to shorten SKILL.md and improve progressive disclosure.

Make the recall-gate feedback loop explicit in the Workflow (e.g., 'if recall drops below threshold, re-quantize or fall back to full precision') rather than only describing the gate.

DimensionReasoningScore

Conciseness

Mostly lean and code/table-driven, but the intro paragraph restates backend roles already captured by the 'Pick a backend' table, and a few phrases ('the strongest privacy posture OpenMed offers') could be trimmed.

4 / 5

Actionability

Fully executable, copy-paste-ready code and CLIs across all three backends — pip install, `python -m openmed.<backend>.convert` with real flags, and Python API calls with concrete parameter values (compute_units, opset, quantize bits).

5 / 5

Workflow Clarity

A numbered 5-step Workflow (Pick → Convert → Quantize → Run → Verify) with explicit validation gates: MLX `--eval-suite` recall certification, 'Verify outputs against the full-precision model before shipping', and hand-off to `evaluating-with-leakage-gates`.

5 / 5

Progressive Disclosure

Well-organized with clear section headers, a comparison table, and one-level-deep pointers (external URLs and source paths like `openmed/mlx/convert.py`); no nested references, though no bundle files exist to offload detail into.

4 / 5

Total

18

/

20

Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, third-person description that concretely states capabilities and provides explicit, natural 'Use when' triggers with synonym coverage. It cleanly separates on-device deployment from hosted/serving/MCP/batch alternatives, giving minimal conflict risk.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'Run OpenMed models fully on-device', 'convert-quantize-run workflows', 'export to ONNX or WebGPU', 'quantize a clinical model to int8/4-bit', 'loading a local MLX artifact through analyze_text', 'OpenMedMLXLanguageModel/generate_text' — giving comprehensive coverage.

5 / 5

Completeness

Explicitly answers both what ('Run OpenMed models fully on-device with the MLX/CoreML/ONNX backends...') and when ('Use when the user wants to deploy OpenMed at the edge...') with concrete trigger phrases.

5 / 5

Trigger Term Quality

Natural trigger phrases a user would say — 'deploy OpenMed at the edge', 'run NER/de-id on Apple Silicon', 'target iPhone/iPad/Mac', 'run with no network', 'pick between MLX/CoreML/ONNX' — with synonym coverage (MLX/CoreML/ONNX, iPhone/iPad/Mac, int8/4-bit).

5 / 5

Distinctiveness Conflict Risk

Clear on-device deployment niche with distinct triggers (MLX/CoreML/ONNX, Apple Silicon, iPhone/iPad/Mac, edge, no network) and explicit routing to sibling skills, minimizing conflict risk.

5 / 5

Total

20

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
maziyarpanahi/openmed
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.