skills.rest
diazMelgarejo
diazMelgarejoCOMMUNITY
Explore all skills by this author →
Website𝕏@Melgarejo_yDiazGitHub

pt-hardware-policy

Validates model-to-hardware affinity using Perpetua-Tools policy before LM Studio dispatch on Windows.

Updated Mar 22, 2026
Search Tags:#windows#model validation#gguf#hermes#lm studio#hardware policy#gpu dispatch
One-click install
npx skills add https://github.com/diazMelgarejo/orama-system --skill pt-hardware-policy-diazmelgarejo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pt-hardware-policy
Source: https://github.com/diazMelgarejo/orama-system/tree/main/bin/orama-system/skills/hermes-harness/commands/pt-hardware-policy
Command: npx skills add https://github.com/diazMelgarejo/orama-system --skill pt-hardware-policy-diazmelgarejo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When Hermes on Windows dispatches models to LM Studio, agents risk guessing hardware affinity rules (NEVER_MAC / NEVER_WIN) from model lists, causing invalid GPU dispatches. This Skill enforces a single canonical policy source from Perpetua-Tools so no local skill invents affinity rules. ## Core Features & Use Cases - Canonical policy validation: Runs the Perpetua-Tools hardware_policy_cli.py to list policy and validate a model against the Windows host before dispatch. - Platform-aware rules: Treats windows_only GGUF models as allowed on the Windows Hermes host and mac_only/MLX models as NEVER_WIN, with the mirror applying on Mac OpenClaw hosts. - Use Case: Before routing a 27B GGUF reasoning model to LM Studio at localhost:1234, an agent runs the policy validation to confirm the model is permitted on Windows, avoiding a failed or misrouted GPU dispatch. ## Quick Start Ask the agent to validate a model against the hardware policy before dispatching it to LM Studio, for example: run the hardware policy check for qwen3.5-27b on Windows.

Frequently Asked Questions about pt-hardware-policy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate a model against hardware policy before LM Studio dispatch?▼

Run the Perpetua-Tools hardware_policy_cli.py with --validate followed by the model name and platform, for example win. On Windows you can also use the platform launcher start.ps1 with the --hardware-policy flag.

What is the difference between NEVER_MAC and NEVER_WIN model rules?▼

NEVER_WIN marks mac_only or MLX models that cannot run on the Windows Hermes host, while NEVER_MAC marks windows_only GGUF models that cannot run on Mac OpenClaw hosts. Both rules come from the same Perpetua-Tools policy file.

Can I infer hardware affinity from the LM Studio /v1/models list?▼

No. The skill explicitly forbids inferring NEVER_MAC or NEVER_WIN rules from model list membership alone. All affinity decisions must be delegated to the Perpetua-Tools hardware_policy module via its CLI or Python import.

Does the hardware policy check require OpenClaw on Windows?▼

No. On Windows Hermes hosts, OpenClaw is optional and typically not installed. The --check-openclaw validation only applies when a Mac OpenClaw openclaw.json configuration is present.

Where is the canonical hardware policy stored?▼

The single source of truth is config/model_hardware_policy.yml in the Perpetua-Tools repository, exposed through src/utils/hardware_policy.py and the scripts/hardware_policy_cli.py CLI. Hermes skills reference it but never duplicate the rules.

Related Skills

Explore More Software Engineering Skills→
eduardtomasekeduardtomasek

grepai-embeddings-lmstudio

Configure LM Studio as the local embedding provider for GrepAI semantic code search.

Community
Basic
afel6afel6

gguf-quantization

Convert and quantize transformer models into GGUF format for CPU inference.

Community
Advanced
epologeeepologee
3

port

Port platform-specific Claude Code skill commands to macOS, Linux, or Windows.

Community
Advanced
SocketDevSocketDev
125

fleet-auditing-gha

Diff GitHub Actions settings against a canonical baseline to surface drift.

Official
Intermediate