mo-qa
Use Momentic's `mo` CLI to run and control Mo, Momentic's cloud autonomous QA agent. Use when starting or continuing Mo sessions, reading their output or status, stopping active work, answering Mo, transferring files, exporting reports, or finding and fixing bugs in a local codeb
Install
npx skills add https://github.com/momentic-ai/skills/tree/main/skills/mo-qa
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install momentic-ai-skills@llmmart
git clone https://github.com/momentic-ai/skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole momentic-ai/skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Run QA with Mo
Mo runs web and mobile QA in the cloud. For repository tests, use
momentic-test or momentic-maintain.
Control Mo with hosted tools or qa, not momentic or momentic-mobile.
Workflow at a glance
Keep the requested interface; in a requested hosted session, report tool or authentication failures without starting a CLI session. Otherwise, use hosted tools, with an existing Connector for private web targets, or the CLI for a new tunnel. If you can't run commands on the machine that reaches the target, have the user start it there, then use hosted tools unless you lack them or the CLI was requested. Read Tunneling and Hosted tools.
Prepare the brief. If effort is unset, ask the user to choose effort and wait. For mobile CLI sessions, follow Mobile bug bash to select a build and start. For web CLI sessions:
session_json=$(qa start "$brief" --effort "$effort") session_id=$(jq -r .sessionId <<<"$session_json") web_url=$(jq -r .webUrl <<<"$session_json")Follow the session with
momentic_mo_get_session, orqa read "$session_id" --from start --timeout 45s --json.For
needs_you, answer withmomentic_mo_message_session, orqa send "$session_id" --wait 45s "<answer>". Pass--question-id "<question-id>"when Mo provides an ID.After
ready, for hosted sessions, read the report withmomentic_mo_get_sessionand summarize findings and uncovered requirements. Callmomentic_mo_stop_sessiononly when the user asks or agrees. For CLI sessions, unless the user wants more testing, runqa report "$session_id" --output <dir>, thenqa stop "$session_id" --subagentseven if export fails, then summarize results; do not use the hosted stop tool. Stop only tunnels you started, after their last session:qa tunnel stop "$tunnel_id".
For read-only requests, inspect without starting, messaging or stopping work. Continue with the same interface. Report the actual stop result, including when the session was already idle.
Prepare the brief
Include the target app/build or URL, goal, requirements, login method, test account label, allowed data, prohibited actions and coverage exclusions. Preserve requested steps; ask about missing scope, access or permission that would change the run. Keep passwords out of the brief.
Read these references when needed:
- Brief when preparing the first brief.
- Installation when
qa versionfails or needs updating. - Authentication when the app needs credentials.
- Tunneling for local or private targets.
- Source context for PRs, local changes or missing repository access.
- Browser settings for non-default browser setup.
- Mobile builds to upload, download, or manage app
builds with
qa assets.
Keep the returned session ID and webUrl; share the link. After an uncertain
start, use startup recovery before retrying.
Follow the session
Read Session control for statuses, questions
and wait behavior. Use blocking waits rather than immediate polls or shell sleeps.
Keep following running and waiting_on_agents. For checking, read the last
reply; if it lasts over a minute, report the unknown outcome without messaging Mo.
For needs_you, answer each pending question using its question ID, or ask the
user. Report blockers. After three consecutive failed reads, report the failures
and unknown outcome, share the link and ask whether to cancel; testing may continue.
Mo manages its testing budget. Poll with 45-second waits; they aren't deadlines.
Monitor time caps with wall time and spend caps with signed-in qa cost.
Hosted MCP doesn't expose cost; the CLI can read hosted session costs.
If spend can't be monitored, tell the user before starting.
At the cap, report coverage cut short. Stop if the user authorized stopping at
the cap; otherwise ask before stopping.
Only ready means the session finished. For incomplete hosted sessions, follow
get_session.nextAction.
Credit findings to Mo; a tester's verified verdict is not a verified session pass.
Message Mo
Relay new findings and status changes as they arrive. Answer user questions from Mo's latest
reply, not a status summary; say when that reply is unavailable. Send messages
only for Mo's questions, user requests or untested areas within the agreed scope.
Don't send continue, finalize or routine progress requests.
Finish
Compare coverage with the brief and ask Mo to cover remaining unblocked areas. Report gaps, findings, the session link and any export directory. Read Reports for evidence export, triage and retests. For fixes, use Mobile bug bash for app builds or Repair loop for web targets. Keep the target and shared data unchanged until testing and reproductions finish.
Files (skills)
-
agents
-
openai.yaml 186 B
Loading…
-
-
references
-
authentication.md 3.3 KB
Loading… -
brief.md 1.2 KB
Loading… -
browser-settings.md 1.9 KB
Loading… -
hosted-tools.md 3.3 KB
Loading… -
installation.md 908 B
Loading… -
mobile-bug-bash.md 3.7 KB
Loading… -
mobile-builds.md 1.7 KB
Loading… -
remediation-loop.md 1.8 KB
Loading… -
reports.md 4.4 KB
Loading… -
session-control.md 3.5 KB
Loading… -
source-context.md 2.8 KB
Loading… -
tunneling.md 3.4 KB
Loading…
-
-
SKILL.md 5.4 KB
--- name: mo-qa description: Run and control Mo web and mobile QA with hosted Momentic tools or the qa CLI, including saved environments, mobile builds, private targets, reports, replies, triage, and retests. --- # Run QA with Mo Mo runs web and mobile QA in the cloud. For repository tests, use `momentic-test` or `momentic-maintain`. Control Mo with hosted tools or `qa`, not `momentic` or `momentic-mobile`. ## Workflow at a glance 1. Keep the requested interface; in a requested hosted session, report tool or authentication failures without starting a CLI session. Otherwise, use hosted tools, with an existing Connector for private web targets, or the CLI for a new tunnel. If you can't run commands on the machine that reaches the target, have the user start it there, then use hosted tools unless you lack them or the CLI was requested. Read [Tunneling](references/tunneling.md) and [Hosted tools](references/hosted-tools.md). 2. Prepare the brief. If effort is unset, ask the user to choose [effort](references/brief.md#effort) and wait. For mobile CLI sessions, follow [Mobile bug bash](references/mobile-bug-bash.md) to select a build and start. For web CLI sessions: ```bash session_json=$(qa start "$brief" --effort "$effort") session_id=$(jq -r .sessionId <<<"$session_json") web_url=$(jq -r .webUrl <<<"$session_json") ``` 3. [Follow the session](#follow-the-session) with `momentic_mo_get_session`, or `qa read "$session_id" --from start --timeout 45s --json`. 4. For `needs_you`, answer with `momentic_mo_message_session`, or `qa send "$session_id" --wait 45s "<answer>"`. Pass `--question-id "<question-id>"` when Mo provides an ID. 5. After `ready`, for hosted sessions, read the report with `momentic_mo_get_session` and summarize findings and uncovered requirements. Call `momentic_mo_stop_session` only when the user asks or agrees. For CLI sessions, unless the user wants more testing, run `qa report "$session_id" --output <dir>`, then `qa stop "$session_id" --subagents` even if export fails, then summarize results; do not use the hosted stop tool. Stop only tunnels you started, after their last session: `qa tunnel stop "$tunnel_id"`. For read-only requests, inspect without starting, messaging or stopping work. Continue with the same interface. Report the actual stop result, including when the session was already idle. ## Prepare the brief Include the target app/build or URL, goal, requirements, login method, test account label, allowed data, prohibited actions and coverage exclusions. Preserve requested steps; ask about missing scope, access or permission that would change the run. Keep passwords out of the brief. Read these references when needed: - [Brief](references/brief.md) when preparing the first brief. - [Installation](references/installation.md) when `qa version` fails or needs updating. - [Authentication](references/authentication.md) when the app needs credentials. - [Tunneling](references/tunneling.md) for local or private targets. - [Source context](references/source-context.md) for PRs, local changes or missing repository access. - [Browser settings](references/browser-settings.md) for non-default browser setup. - [Mobile builds](references/mobile-builds.md) to upload, download, or manage app builds with `qa assets`. Keep the returned session ID and `webUrl`; share the link. After an uncertain start, use [startup recovery](references/session-control.md#recover-startup) before retrying. ## Follow the session Read [Session control](references/session-control.md) for statuses, questions and wait behavior. Use blocking waits rather than immediate polls or shell sleeps. Keep following `running` and `waiting_on_agents`. For `checking`, read the last reply; if it lasts over a minute, report the unknown outcome without messaging Mo. For `needs_you`, answer each pending question using its question ID, or ask the user. Report blockers. After three consecutive failed reads, report the failures and unknown outcome, share the link and ask whether to cancel; testing may continue. Mo manages its testing budget. Poll with 45-second waits; they aren't deadlines. Monitor time caps with wall time and spend caps with signed-in `qa cost`. Hosted MCP doesn't expose cost; the CLI can read hosted session costs. If spend can't be monitored, tell the user before starting. At the cap, report coverage cut short. Stop if the user authorized stopping at the cap; otherwise ask before stopping. Only `ready` means the session finished. For incomplete hosted sessions, follow `get_session.nextAction`. Credit findings to Mo; a tester's verified verdict is not a verified session pass. ## Message Mo Relay new findings and status changes as they arrive. Answer user questions from Mo's latest reply, not a status summary; say when that reply is unavailable. Send messages only for Mo's questions, user requests or untested areas within the agreed scope. Don't send `continue`, `finalize` or routine progress requests. ## Finish Compare coverage with the brief and ask Mo to cover remaining unblocked areas. Report gaps, findings, the session link and any export directory. Read [Reports](references/reports.md) for evidence export, triage and retests. For fixes, use [Mobile bug bash](references/mobile-bug-bash.md#finish-and-recheck) for app builds or [Repair loop](references/remediation-loop.md) for web targets. Keep the target and shared data unchanged until testing and reproductions finish.
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.