Look twice.Find the gem.

AI agents and MCP servers, each published with its source and what the checks found.

Marketplace

  • Everything
  • AI agents
  • Apps
  • MCP servers
  • Templates
  • What people want
  • What changed this week
  • The verification standard
  • The ooruby Index
  • Servers that publish no source
  • Reliability guides
  • What the catalogue holds
  • Sell here

Our library

  • Everything, in one place
  • Guides
  • Glossary
  • Calculators
  • Checklists and cheat sheets
  • Community

ooruby

  • Home
  • For teams
  • Site status
  • Company projects
  • RSS feed

Verification records what our published tests found on a specific version at a specific date. It is not a warranty, and it does not certify that software is free of defects.

AI agentsAppsMCP serversTemplatesWantedCommunityOur library
Sign inSell
Catalogue/Agents/Workspace Probe

Workspace Probe

metaharness · Agent

Workspace Probe is the evaluation bridge for the workspace-lens package. It takes the interpretability receipts that lens produces for each model decision and computes a workspace probe score, along with mean drift and flag, critical and clean rates. It also grades a baseline against a mutant and rejects a prompt or policy mutation that improves the final answer at the cost of the model's internal process. It works as a library and as a command line with JSON in and out, and needs a fitted lens produced separately.

Written by ooruby from the README of @metaharness/workspace-probe 0.1.1.

Not claimed by its maker.

Why it's useful

It flags prompt or policy changes that improve the final answer while making the model's internal process less steady.

What you could do with it

  1. 1Score a set of receipts to see what share of decisions kept a clean workspace and how much it drifted.
  2. 2Compare baseline and mutant receipts and get a keep or veto verdict with the reasons behind it.
  3. 3Run the command line on JSON files to read a fitted lens's metadata or the workspace tokens for each activation.

Written from its documentation: what it says it can do, not what our checks found.

WPagentDeveloper tools

Maintenance

Not measured

No release date is on record for this package yet. The nightly publisher check records one as it reaches each npm package, about once every ten nights.

Release activity only: it says nothing about quality or safety, and a finished small package can be fine without releases. How it is measured

Not yet tested
Licence and provenance
finding
no LICENCE file at the root (.)
Data handling disclosed
finding
no statement of what it reads, writes or sends (README.md)

Published because it was found. A finding is information about where the limits are, not a verdict that the software is unsafe.

What this does not cover

  • ·Everything. Treat it as you would software from anywhere else.

Receipt history: every signed receipt for every version of this listing, and what changed between them.

In its maker's words

Evaluation + Darwin-Mode bridge for @metaharness/workspace-lens: turn Jacobian-Lens interpretability receipts into a flywheel-consumable workspace_probe score, and reject structurally-brittle prompt/policy mutations (final answer up, workspace grip down).

Compare with

  • Acpxside by side
  • Fallowside by side
  • Vanguard Frontier Agenticside by side
Compare all 4

How far this has been checked

Static scanned

What it establishes
Its published package, or for a hosted server its source repository at a recorded commit, was read file by file, without running it, by every check in our current scan.
What it does not
How it behaves when it runs, or anything the scan does not read yet: several rubric checks, the repository's history, and compiled code in folders named dist or build. For a hosted server, that the endpoint runs the code that was read. The ooruby Index sets out exactly what was read.
Exactly what was read
Version 0.1.1, the file with digest sha512-kHL7T16476a5obVunbZ2NYqK7YY/1qtZTaQC1KyXmFh/7dRNselqGAilawspuD86F9U1H/LMCDvchDj3Vh5duw==, as the registry publishes it. A published version cannot be replaced, so this names the same bytes for anyone who checks.
Authentication
Runs locally. No hosted endpoint is listed for this package. You run it on your own machine, so there is no endpoint of ours to authenticate against.
How the rungs work

Adding it

Run it with:

npx -y @metaharness/workspace-probe@0.1.1

Pinned to 0.1.1, which is the version the findings above were found in. Drop the version to take whatever is newest, and the report on this page stops describing what you installed.

Where this came from

Registry
npm
Package
@metaharness/workspace-probe
Version read
0.1.1
Resolved
2026-10-10
Open the source(opens in a new tab)

No build provenance published. The repository above is the one the publisher declared, and nothing links it to the package you would install. That is not a mark against this listing, since most packages are published this way, but it is a check nobody can run.

The badge, if this is your listing

It renders the current rung (static scanned) and links back here, where what that does and does not establish is one click away. It updates itself as the evidence deepens.

[![ooruby: static scanned](https://ooruby.com/api/badge/metaharness-workspace-probe)](https://ooruby.com/market/metaharness-workspace-probe)

About this listing

Kind
Agent
Category
Developer tools
Pricing
Free
Sandbox
No
Hosted
No
Updated
2026-10-10
Price
Free
Get it freeVisit maker
Rating
no reviews yet
0 want this so far
Sign in to vote

Sign in to watch this listing and hear when its version, rung or findings change.

More from metaharness

See the storefront
DAagentDeveloper tools
4 findingsStatic scanned
Run one evolve command on a repository and it mutates one surface at a time across several generations. It ranks the variants on a leaderboard and writes the population and lineage to a work folder in the repo.

Darwin

metaharness
Free

Evolves the planner, reviewer and retry policies around your model, testing each change in a sandbox.

347.3k installs/moMIT
REagentDeveloper tools
Clean scanStatic scanned
Attacks cover prompt injection, tool misuse, data leakage and cost amplification, and a model judge scores severity. The real judge needs an OpenRouter API key, and a mock judge exists for tests.

Redblue

metaharness
Free

Attacks an AI agent you own with adversarial models, judges what broke, patches it and retests.

163.1k installs/moMIT
WEagentDeveloper tools
Clean scanStatic scanned
Exports SFT and DPO sets in standard formats such as OpenAI chat JSONL and TRL, with tool calls kept, plus a GPU training plan. It drops runs that overlap your eval set or tampered with tests.

Weight Eft

metaharness
Free

Turns your coding agent's past runs into LoRA training data for a cheaper open model such as Qwen.

147.3k installs/moMIT

Similar agents

All agents
OCagentDeveloper tools
3 findingsStatic scanned
Sign in with a ChatGPT account to use it under a Plus, Pro, Business, Edu or Enterprise plan, or use an API key with some extra setup. Separate options exist for code editors, a desktop app and the cloud-based Codex Web.

Openai Codex

openai
Free

Runs OpenAI's Codex coding agent locally on your computer from the command line.

71.3M installs/moApache-2.0
SKagentDeveloper tools
Clean scanStatic scanned
Skills come from GitHub repositories, given as owner and repository or as a full address, and you can try one without installing it. Updates can cover global skills, project skills or a single named skill.

Skills

rauchg
Free

Installs, finds, updates and removes agent skills across many coding agents from one command line.

38.3M installs/moMIT
ADagentDeveloper tools
Clean scanStatic scanned
It checks environment variables and reports the agent's ID, display name and session ID where one exists, with an opt-in process-tree check. A command line check answers by exit code.

Agent CLI Detector

ide
Free

Lets your JavaScript command line tool tell whether a coding agent is running it, and which one.

14M installs/moMIT

Verification records what our published tests found on a specific version at a specific date. It is not a warranty, and it does not certify that software is free of defects.