Detect harness problems
Checks committed harness files against repository paths and configuration, identifying conflicts, missing-file references, patterns that match no files, and over-broad permissions with evidence and detector coverage.
tetriz-harness
tetriz-harness is an open-source developer tool that scans and improves the rules, skills, hooks, agents, and other configuration used by AI coding tools. It helps engineering teams review harness recommendations and distribute approved changes to selected developers or teams, including manually managed test/control rollouts.
AI coding harness recommendations and controlled distribution
tetriz-harness is an open-source developer tool for evaluating and improving the harnesses that guide AI coding agents. Its harness model includes rules, skills, agents, commands, hooks, and MCP configuration for tools such as Claude Code, Codex, Cursor, and Kiro.
The repository combines a repository-aware scanner and recommendation engine with a control plane for distributing approved changes. Teams can inspect detected problems, use coding and organizational signals to derive additional recommendations, review findings, then target selected artifacts to developers or groups. Audience targeting also supports manually managed test/control rollouts, while outcome analysis remains external to the control plane.
Core capabilities
Derive recommendations from coding signals
Uses organizational session analyses, code-review themes, and security signals to propose new rules, agents, and hooks; this workflow depends on populated signal data, the bundled tetriz-py service, and a model key.
Use inventory and usage evidence
Can incorporate personal harness inventories and invocation records supplied through an ingestion layer for usage-informed checks; an on-disk checkout scan cannot infer those inputs by itself.
Review and approve changes
Persists findings and recommendations so teams can review them and explicitly choose which changes to adopt before publication.
Distribute approved artifacts
Provides a Go server library, reference host with an admin UI, harness-agent CLI, and Node library for distributing approved harness artifacts.
Target and monitor rollouts
Targets rules, skills, hooks, and agents to named people, teams, or the whole organization, installs them in supported user configuration locations or known repository checkouts, and reports per-person installation status.
Practical use cases
Audit an existing repository harness
Run the DETECTED scan against committed harness files to find stale references, ineffective file patterns, conflicts, and permissions that are broader than intended before changing the harness.
Turn team signals into proposed artifacts
Combine session analyses, code-review themes, and security signals to identify candidate rules, agents, or hooks that address recurring engineering patterns.
Manage a controlled team rollout
Review and publish only selected recommendations, then deliver them to named developers, teams, or the full organization while checking installation status.
Run a manual harness experiment
Create disjoint test and control groups, target a harness change only to the test group, and compare coding outcomes using an external session or review analysis system.
Enrich analysis with local agent evidence
Use the standalone Agent Desk discovery and session-extraction packages through an adapter to supply harness inventory and observed activity to the recommendation workflow.
Pros and Cons
Pros
- Combines repository-aware static checks with recommendations based on coding, review, security, inventory, and usage signals.
- Keeps adoption explicit by separating recommendation review from publication and distribution.
- Supports targeting by person, team, or organization and reports installation status for targeted users.
- Available as composable Go, Node, CLI, server, and admin UI components rather than only as a hosted interface.
Cons
- Signal-driven recommendations require additional data sources, an ingestion adapter, the bundled tetriz-py service, and a model key.
- The repository documents a manually managed rollout process rather than cohort randomization, built-in experiment analysis, or automatic version restoration.
- The Agent Desk pairing does not directly import its JSON or SQLite stores; an adapter is required.
FAQ
What is tetriz-harness?
It is an open-source repository that scans AI coding harness files, generates recommendations from repository and organizational signals, and distributes approved artifacts through a control plane.
What inputs and setup are needed?
The DETECTED scan can inspect committed harness files against repository paths and configuration without making a model call. Recommendation workflows that use organizational signals require populated signal tables, the bundled tetriz-py service, and a model key.
How are approved changes distributed?
The control plane can target approved artifacts to named people, teams, or the whole organization, place them in tool user configuration or known repository checkouts, and report installation status.
Can it support a test and control rollout?
Yes. Audience targeting can deliver a selected artifact to a test group while leaving a separate control group untargeted. The system reports delivery state, but outcome measurement and experiment analysis remain in the user's own analysis system.
What are the documented rollout limitations?
The repository states that it does not randomize cohorts, calculate experiment results, automatically restore an earlier harness version, or guarantee that untargeted users lack equivalent files from another source.
Quick Facts
- Category
- Developer tool
- Primary focus
- AI coding harness analysis and distribution
- Repository
- GitHub / tetriz-ai/tetriz-harness
- Harness artifacts
- Rules, skills, agents, commands, hooks, and MCP configuration
- Deployment components
- Go libraries and server, harness-agent CLI, Node library, reference admin UI
- Pricing
- No product pricing information is provided in the available source
Alternativas a tetriz-harness
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads inside Firecracker micro-VMs. It is designed for workflows that need machine-level isolation, private networking between sandboxes, and programmatic control through SDK, CLI, or MCP.
GitBot
GitBot is a local browser-based workspace for creating reusable coding bots powered by Claude Code, Codex, or OpenCode. It helps developers run those agents across repository folders with persistent threads and configurable permissions.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repositories and verifies changes with the real compiler, debugger, sanitizers, and tests before showing a diff. It offers a free tier plus paid plans, with editor connectors and zero-retention handling described in the source.
hob
hob is an independent workspace for coding agents that keeps agent sessions, terminals, history, and follow-up work organized around the tools and providers you already use. It is aimed at developers who want local control over routing, history, and workspace structure rather than a bundled model stack.
Manta AI
Manta AI is an autonomous web app testing tool for teams that want to map application behavior, catch regressions, and generate tests without writing scripts or maintaining selectors. It works from a URL and supports plain-English test flows, run results with screenshots, and scheduled or deployment-triggered checks.
SonOf
SonOf connects to your repo and PM tool, audits the codebase and surrounding product context, and turns approved work into shipped tickets with senior engineering review. It is aimed at founders and engineering leaders who need backlog help without hiring a full team immediately.