aide-inhouse-only-policy-hook SKILL.md skillA

aide-inhouse-only-policy-hook from anonymousnomad/aide-sovereign-workbench is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add anonymousnomad/aide-sovereign-workbench/aide-inhouse-only-policy-hook@git:20260829.3450c8e

Pins git:20260829.3450c8e in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
anonymousnomad/aide-sovereign-workbench
Path
skills/packs/aide-inhouse-only-policy-hook/SKILL.md
Versions recorded
1
Latest version
git:20260829.3450c8e, committed Aug 29, 2026
Size
8,111 bytes
Repository stars
53
License
Apache-2.0

The doctrine that AIDE's desktop-control act() MUST route the policy inference through the in-house model (Cipher v1, 4B) — never a cloud model, never a BYOK provider. Why this is the differentiator (a small model fine-tuned on a verified desktop-control trajectory corpus is more reliable on that exact task than a giant cloud model with no domain adaptation), how to gate it in routes/desktop.ts and the agent's desktop_action tool, and how the operator-overrides-the-safety-rails exception works…

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

# Desktop Control — In-House Model Only

Born 2026-08-28 from the operator's directive: "as of right now the
only model we're going to allow to be used for desktop control is
our in-house model because we have specifically fine tuned for this
kind of workflow." This is the doctrine for that decision and the
mechanics that enforce it.

## Why this is the differentiator

The competitive claim is **not** "AIDE can drive a desktop" — every
cloud agent in 2026 can do that. The claim is **"AIDE's house model
drives the desktop, locally, with operator-owned safety rails."** Three
reasons in priority order:

1. **Domain adaptation beats scale, on narrow tasks.** Cipher v1 is
   a 4B model that we have specifically fine-tuned on verified
   desktop-control trajectories (1,098 rows per AGENT_NOTES
   2026-08-26, 90/90 generator battery). A giant cloud model with
   no desktop-specific training has a higher loss on the exact
   question "given this Windows snapshot and this grants manifest,
   emit the correct desktop_action DSL". A 4B model that's seen
   the task 1,000 times beats a 1T model that's seen it 0 times.

2. **Latency + all-local loop.** The desktop act() cycle is
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add anonymousnomad/aide-sovereign-workbench/aide-inhouse-only-policy-hook@sha256:489cc1dd7724fe00

The line at the top of this page pins the label git:20260829.3450c8e, 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.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_orw7qf3ad5uxwpnv.svg)](https://markdownregistry.com/a/art_orw7qf3ad5uxwpnv)

4 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 1 version of this file, committed on Aug 29, 2026.

versioncommittedcommitsizeaudit
git:20260829.3450c8e latest2026-08-29 3450c8e 8,111 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (8111 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

GitHub

anonymousnomad/aide-sovereign-workbench · 53 stars · license Apache-2.0 · pushed 2026-10-26 · branch covert-production

API

GET https://markdownregistry.com/api/v1/artifacts/art_orw7qf3ad5uxwpnv
GET https://markdownregistry.com/api/v1/resolve?ref=anonymousnomad/aide-sovereign-workbench/aide-inhouse-only-policy-hook
GET https://markdownregistry.com/api/v1/blob/489cc1dd7724fe00075e6645835e999441418018ba546cfe950ffa3f7a8b7a93

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.

More from anonymousnomad/aide-sovereign-workbench

AGENTS.md AGENTS.md file
anonymousnomad/aide-sovereign-workbench · AGENTS.md
git:20260911.2b7569f · audit A · 53 stars
audit-probe-discipline SKILL.md skill
anonymousnomad/aide-sovereign-workbench · docs/phase0/skills/audit-probe-discipline/SKILL.md · Local Phase-0 diagnostic procedure for quoting, stale runner references, and restricted Windows inventory queries.
git:20260912.9e65c9f · audit A · 53 stars
llms.txt llms.txt file
anonymousnomad/aide-sovereign-workbench · llms.txt
git:20260916.e7b3bfe · audit A · 53 stars
PC management SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/PC management/SKILL.md · Owner's manual for THIS machine (ASUS GL703GM — i7-8750H / GTX 1060 / 16GB RAM / 120Hz). Use whenever helping with…
git:20260823.df7cbde · audit A · 53 stars
adapter-serve-routing SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/adapter-serve-routing/SKILL.md · Define and verify deterministic format-conditional adapter routing at inference. Use whenever a shared base model…
git:20260823.df7cbde · audit A · 53 stars
agent-notes SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/agent-notes/SKILL.md · Agent Persistence System — the mandatory project journal. Use at the START of every session and every context…
git:20260823.df7cbde · audit A · 53 stars
aid-capability-battery-thinking-models SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-capability-battery-thinking-models/SKILL.md
git:20260908.e9ac98d · audit C · 53 stars
aid-cipher-4b-fine-tune-pipeline SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-cipher-4b-fine-tune-pipeline/SKILL.md · End-to-end recipe for fine-tuning the AIDE in-house cipher-4b model: fix the broken venv, install the ML stack…
git:20260908.e9ac98d · audit B · 53 stars
aid-closed-loop-on-by-default SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-closed-loop-on-by-default/SKILL.md · Wire AIDE's self-improvement loop (`scripts/selfimprove.mjs`) into the daemon lifecycle so it runs continuously, not…
git:20260908.e9ac98d · audit A · 53 stars
aid-closed-loop-self-improvement SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-closed-loop-self-improvement/SKILL.md
git:20260908.e9ac98d · audit C · 53 stars
aid-dap-real-debugpy-fixture SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-dap-real-debugpy-fixture/SKILL.md · Wire in a REAL debugpy fixture for the AIDE DAP test batteries. The tests/arch/dap-contract.test.ts and…
git:20260908.e9ac98d · audit A · 53 stars
aid-double-check-everything SKILL.md skill
anonymousnomad/aide-sovereign-workbench · skills/packs/aid-double-check-everything/SKILL.md
git:20260908.e9ac98d · audit C · 53 stars

Every file in anonymousnomad/aide-sovereign-workbench

Browse by kind, by grade A, or by owner.