optimize-local-model-compression SKILL.md skillA
optimize-local-model-compression from understudylabs/orchestra-agent-tools is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add understudylabs/orchestra-agent-tools/optimize-local-model-compression@git:20261027.533318ePins git:20261027.533318e in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- understudylabs/orchestra-agent-tools
- Path
skills/optimize-local-model-compression/SKILL.md- Versions recorded
- 3
- Latest version
- git:20261027.533318e, committed Oct 27, 2026
- First version recorded
- Jun 26, 2026
- Size
- 8,883 bytes
- Repository stars
- 19
- License
- MIT
Use when a developer needs to compress a local model for tool-calling workloads — "which quantization should I use", "why does my model emit broken JSON", "how do I get better tool-call fidelity from a 4-bit model", "OptiQ vs QAT vs naive", "the model stopped calling tools after quantization". Covers layer-aware sensitivity-driven compression, QAT group-size matching, and outcome-optimized calibration for structured output.
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# Optimize Local Model Compression Compression (quantization to 4-bit) is mandatory for local deployment — it is the difference between a 9.5 GB model and a 3.6 GB model, between 23 tok/s and 45 tok/s. But standard compression methods optimize for general text quality (perplexity, MMLU) and **silently destroy tool-call fidelity**. This skill teaches the method we invented to fix that: outcome-optimized, layer-aware compression that protects the circuits responsible for structured output. **This is not theory.** Every recommendation below is backed by measured results on Gemma 4 E2B through 31B, on a 103-row tool-call benchmark and a 14-task multi-turn agent board, on Apple Silicon (M5 Max, 128 GB), at zero API cost. ## When to use - The user's compressed model emits broken JSON, fails to call tools, or scores lower on tool-calling tasks than the hosted version. - The user is choosing between quantization methods (naive 4-bit, QAT, OptiQ, or a custom conversion). - The user wants to convert a model and is asking which settings matter. - A local eval shows a model "can't do tool calls" but the hosted version can. ## Safety Gates …
Read the whole file at its exact version.
How to install
mdr add understudylabs/orchestra-agent-tools/optimize-local-model-compression@sha256:bb917ed046cd9910The line at the top of this page pins the label git:20261027.533318e, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_dveqysplicpt5uul)
1 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 3 versions of this file, between Jun 26, 2026 and Oct 27, 2026, about one every 62 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20261027.533318e latest | 2026-10-27 | 533318e | 8,883 B | A | view · diff |
| git:20260626.49d4593 | 2026-06-26 | 49d4593 | 8,531 B | A | view · diff |
| git:20260626.caba8a7 | 2026-06-26 | caba8a7 | 8,884 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (8883 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
understudylabs/orchestra-agent-tools · 19 stars · license MIT · pushed 2026-11-05 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_dveqysplicpt5uul GET https://markdownregistry.com/api/v1/resolve?ref=understudylabs/orchestra-agent-tools/optimize-local-model-compression GET https://markdownregistry.com/api/v1/blob/bb917ed046cd9910d3a1dd529a8dd9c09a5e10c53e53a75d2544339c2c68af9c
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.