momentic-maintain Skill for Claude Code: What It Does and How to Install It

Makes Claude find why Momentic tests fail, classify the cause and repair the tests, including with momentic ai triage.

Installs: 27.3K; on skills.sh
Stars: 16; on GitHub
Updated: Oct 2026; last repo commit

This skill makes Claude diagnose and repair failing Momentic tests. It answers questions like "what happened on my branch" and "why is main red", re-classifies runs with momentic ai classify, and repairs tests with momentic ai triage.

Triage can leave changes locally, write a patch, commit or open a pull request, so set the delivery choice yourself. It needs a Momentic API key.

Good for

Developers asking why a branch is red
DevX and on-call engineers de-flaking quarantined or retry-heavy tests

Install momentic-maintain

Claude Code

Copy this prompt.

Install the momentic-maintain skill in Claude Code. Run this command:

/plugin marketplace add momentic-ai/skills
/plugin install momentic@momentic

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

The plugin route installs all 6 skills (skills/*/SKILL.md in the repo tree) and registers the momentic and momentic-mobile MCP servers from .mcp.json. The skill runs npx momentic ai commands, calls the Momentic API with MOMENTIC_API_KEY and can start your app, so there is no Claude app upload route. Sign in once with the login command in the last method, or set MOMENTIC_API_KEY.

From the project README. Source.

What's in the skill

When it loads

Diagnose, classify, triage, and repair failing Momentic tests with MCP run tools, the Momentic CLI, and manual run artifacts. Use when a developer asks what happened on a branch, DevX or on-call asks why main is red, or the user wants to inspect classifications, de-flake quarantined or recovered tests, reduce retries, re-classify runs, run AI triage, or repair failures.

Sections in SKILL.md

Common workflows
Maintenance workflow
Inspect the existing classification
Helpful MCP tools
Investigation workflow
Evidence standards
Run AI classification
Prepare triage
Run AI triage
Download run artifacts manually
Background
Test run result structure
Element locators
Module caching
File uploads
Using past runs

Key facts

Author: Momentic (vendor's own repo)
License: None stated: the repo has no LICENSE file and its plugin manifests list no license
Installs on skills.sh: 27,324 (Nov 2, 2026)
GitHub stars: 16 for the whole repo (Nov 2, 2026)
Last repo commit: Oct 28, 2026
Audits on skills.sh: Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass
What it includes: 1 file: SKILL.md (from the repo tree)
Needs: MOMENTIC_API_KEY for the API calls, a Momentic project, and the Momentic MCP server for run tools
Part of: The Momentic plugin: 6 skills (skills/*/SKILL.md in the repo tree) and 2 local MCP servers listed in .mcp.json
Upstream sync: The README says a bot copies most files in skills/ from an upstream source, so edits made in the repo are overwritten

What it includes

What happened on my branch, why main is red, de-flake and cut retries
Uses momentic_list_runs with exact branch names, time bounds and filters.
Try: “Why is main red? Group the failures by cause.”
Network · Common workflows
curl run metadata from the Momentic API with MOMENTIC_API_KEY
Uses api.momentic.ai unless MOMENTIC_SERVER is set, and jq to read the result.
Runs code · Inspect saved classification
get_run, list_runs, quarantine_list, get_step_result, get_test_steps_for_run
Plus momentic_submit_result_classification, which saves a manual verdict only when you ask.
Network · MCP run tools
Runs the classifier on a run, run group or git commit
Add --save only to replace the saved verdict. --interactive keeps the chat open for one run.
Runs code · npx momentic ai classify
Build the app, start services, check readiness, use --parallel 1
Can use --start and --wait-on so triage starts your app, for example with pnpm start.
Runs code · Prepare triage
Groups failures, edits tests, replays them and keeps verified repairs
--on-heal-success picks nothing, patch, a pull request or a commit. --dry-run --no-save previews the grouping.
Runs code · npx momentic ai triage
Export endpoint gives signed ZIP links per attempt; unzip to a temp folder
For when MCP output is missing. Remove the temp folder afterwards.
Runs code · Download run artifacts
The same 8 category ids, recoverability and confidence as classification
Exactly one category per run.
Eight categories

From the official docs

Permissions

High risk

This skill can edit tests and, depending on one flag, commit or open a pull request for you, and it sends your Momentic API key to the Momentic API. Our advice: set --on-heal-success to nothing or patch, review every diff, and keep the key out of logs.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

AI triage can commit or open a pull request
SKILL.md says momentic ai triage can leave changes locally, produce a patch, commit or open a pull request, and tells Claude to set --on-heal-success explicitly. SKILL.md says to pass --yes only after the environment and delivery choice are confirmed, so dependency prompts do not stall an autonomous run. Our advice: use nothing or patch unless you want a PR. github.com
It sends your API key to Momentic from curl
The metadata and export calls send MOMENTIC_API_KEY as a Bearer token to api.momentic.ai, or to MOMENTIC_SERVER if set. SKILL.md says not to echo, log or commit the key. github.com
Triage can start your app
The example passes --start "pnpm start" and a --wait-on URL so triage launches your app. SKILL.md says not to copy the examples blindly and to find the real commands from your repo and CI. github.com
Downloaded run data may be sensitive
SKILL.md says to treat downloaded metadata and signed artifact URLs as sensitive run data, not to commit them, and to remove the temporary folder after the investigation. github.com
classify --save replaces the saved verdict
SKILL.md says to add --save to momentic ai classify only when the new verdict should replace the saved classification. github.com
The plugin route adds two local MCP servers
The plugin route installs the repo's .mcp.json, which starts momentic and momentic-mobile with npx -y. The repo tree has no hooks. Two skills (mo-qa and momentic-spec) bundle an agents/openai.yaml that holds a display name, a short description and a default prompt. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	SKILL.md describes no read-only mode.
Talks to	api.momentic.ai (default MOMENTIC_SERVER_URL for the curl calls). app.momentic.ai (run links it cites in answers). Short-lived signed artifact URLs from the export endpoint
Sign-in and keys	The curl commands send MOMENTIC_API_KEY as a Bearer token. SKILL.md says not to echo, log or commit the key.
Scopes	Reads Momentic run results through MCP tools and the Momentic API Runs npx momentic ai classify and npx momentic ai triage in your project Edits Momentic test files and can leave them local, write a patch, commit or open a pull request Downloads run artifacts to a temporary folder and unzips them Can start your app with the start command you give triage

Questions

What does the momentic-maintain skill do?

It makes Claude diagnose failing Momentic tests: list and inspect runs, re-classify failures with momentic ai classify, and repair tests with momentic ai triage. It also covers quarantined and retry-heavy tests.

How do I install it?

Use the install box at the top of this page. The plugin route installs all six Momentic skills. You also need MOMENTIC_API_KEY set for the API calls and the Momentic CLI available through npx.

Do I need a Momentic account?

It needs a Momentic API key. SKILL.md reads MOMENTIC_API_KEY for its curl calls, and the README says to run npx momentic login once or set the key.

Is it an official Momentic skill?

Yes. It lives in Momentic's own repo, momentic-ai/skills, and the plugin manifest names Momentic as author. The README says a bot copies most skill files from an upstream source, so Momentic asks outside contributors to open an issue instead of a pull request.

Is it safe?

It runs npx momentic ai triage, which can edit tests and commit or open a pull request depending on --on-heal-success. It also sends your Momentic API key to the Momentic API and downloads run data. Our advice: set the delivery option to nothing or patch and review every diff.

Does it work in Cursor or Codex?

The README documents a Cursor plugin from the Cursor Marketplace and a Devin plugin. It also says install-mcp writes the MCP server entry for Claude Code, Cursor, VS Code, Codex, Windsurf and other clients. The skill files are plain SKILL.md, so the skills CLI can install them in other agents.

Can it open a pull request?

It can, through momentic ai triage. SKILL.md lists nothing, patch and pull-request or commit options for --on-heal-success and says to use a pull-request or commit option only when you asked for that delivery.

How is it different from momentic-result-classification?

momentic-maintain adds branch and main-red workflows, momentic ai classify, momentic ai triage repairs and manual artifact downloads. momentic-result-classification only investigates and labels one run with MCP tools.

Information

updated Nov 2, 2026

Repository: momentic-ai/skills
Folder: skills/momentic-maintain

momentic-ai/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5

README · github.com
.mcp.json · github.com
marketplace.json · github.com
momentic-maintain SKILL.md · github.com
momentic-result-classification/SKILL.md · github.com