Guide · Agent skill
How to use the Agent QA Result Triage agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Agent QA Result Triage is an agent skill that helps an AI agent work out why an automated Agent QA test run failed, using the run's recorded evidence rather than guesses. It is written for QA engineers and developers who have a failed run and need to know whether the fault lies in the test, the product, the browser or mobile runtime, or the infrastructure. This guide explains what the skill does, how to install it, how to ask for a triage, what it needs and what it will not do.
Agent QA Result Triagesickn33/agentic-awesome-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the Agent QA Result Triage skill as a ZIP from the page where you found this guide.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into the project's .claude/skills/ folder to limit it to one project.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the ZIP.
- 4Connect the Agent QA MCP tools. Set up the Agent QA MCP server so the agent can call tools such as agent_qa_get_run and agent_qa_classify_failure, or plan a fallback to the dashboard REST APIs or Agent QA CLI.
- 5Ask for a triage. Name the failed run and ask why it failed, for example by asking for the logs and screenshots to be checked.
- 6Review the evidence. Check the category, confidence and evidence listed in the result, and note any missing artifacts that lower confidence.
- 7Hand off any fix. If the triage points to a code change, use the separate agent-qa-debug-fix skill for the repair.
What does the skill make the agent do?
The skill sets a fixed process for investigating one failed or interrupted Agent QA run. The agent first checks the run's status, suite context, steps and attempts. It then fetches the artifacts, steps and logs before deciding anything.
It must choose exactly one failure category from a fixed list. It then returns a short triage result: the category, a confidence level, the evidence it relied on, the likely fix area and a next action. If the failure needs a code change, the skill tells the agent to hand off to the separate agent-qa-debug-fix skill once triage is finished.
- Reads run status, steps and attempts before classifying
- Fetches artifacts, steps, logs and execution logs as evidence
- Uses one category: timeout, appium_startup, browser_disconnect, element_not_found, assertion_failure, hook_failure, infrastructure or unknown_failure
- Reports confidence, evidence, likely owner and next action
- Can compare several recent related runs to find a repeating pattern
Install the Agent QA Result Triage skill
Download the skill as a ZIP from the page where you found this guide. Then choose one of two installation routes.
For a personal setup, unzip the folder into your user skills directory at ~/.claude/skills/. To limit the skill to one project, unzip it into that project's .claude/skills/ folder instead.
If you use the Claude apps, you can upload the ZIP instead. Open Settings, go to Capabilities, then Skills, and add the skill there.
- Personal: ~/.claude/skills/
- Single project: the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Connect the Agent QA tools the skill needs
The skill does not work on its own. It relies on the Agent QA MCP server to supply run data. The source names tools such as agent_qa_get_run, agent_qa_get_run_steps, agent_qa_get_run_logs, agent_qa_get_run_execution_logs, agent_qa_get_run_artifact and agent_qa_classify_failure. The source does not give setup steps for the MCP server, so follow the Agent QA documentation for that.
If MCP is unavailable, the skill can fall back to the dashboard REST APIs or the Agent QA CLI. In that case the agent must state which evidence it could not get. You need an Agent QA account or access to the run data the service stores.
How do I ask for a triage?
Name the run and say what you want. The skill is triggered by requests about failed or interrupted Agent QA runs, their artifacts or logs, and recurring failures. You do not need special syntax.
Good requests give the run identifier and the question. The agent will then fetch the evidence before answering. Requests that ask for a code fix are better sent as a separate request after the triage.
- Why did Agent QA run 4812 fail? Check the logs and screenshots.
- Triage this interrupted run and tell me who should look at it.
- Compare the last five failed runs in this suite and tell me if the same category keeps appearing.
- Does this failure look like a test problem, a product problem or infrastructure?
Read the triage output
The source gives an example of the output format. Each result has a category, a confidence value, evidence items, a likely fix area and a next action. The evidence must come from artifacts, logs or steps that the tools actually returned.
The skill tells the agent to lower confidence when artifacts are missing, such as screenshots, DOM or accessibility context, device logs or earlier runs. Treat a low-confidence result as a lead to check, not a verdict.
- Category: one of the eight fixed values
- Confidence: how well the retained evidence supports the category
- Evidence: quoted or summarised artifact, log or step results
- Likely fix area: for example the test definition or changed product UI
- Next action: the next evidence-backed check
{"category": "element_not_found", "confidence": "high", "evidence": ["Step 4 could not resolve the described checkout button"], "likely_fix_area": "test definition or changed product UI", "next_action": "Inspect the captured UI context, then compare the current checkout screen"}Safety, limits and what the skill will not do
The skill uses the network, because it fetches run data from Agent QA. It does not run scripts. It does not modify tests or application code. The source says the skill redacts credentials, session tokens, personal data and unrelated application content from its report.
The category is the most likely failure surface, not proof of root cause. Check the licence before commercial use: the listing shows MIT, but the skill's own metadata names FSL-1.1-ALv2 and links to the LICENSE file in the Agent QA repository, so read that file to confirm the terms that apply.
- Uses network: yes, to reach Agent QA run data
- Runs scripts: no
- Does not modify code or tests
- Redacts credentials, tokens and personal data from reports
- Confidence drops when retained artifacts or logs are missing
What people use it for
- Explaining why an overnight Agent QA run timed out or stopped partway through
- Deciding whether a failed checkout test is a broken test, a changed product screen or an infrastructure fault
- Comparing several failed runs in one suite to see whether one category keeps recurring
- Checking whether a mobile run failed because Appium could not start a session
- Producing a short triage note with evidence to pass to the team that owns the failing area
Frequently asked questions
Is the Agent QA Result Triage skill free?
The listing shows an MIT licence, but the skill's metadata names FSL-1.1-ALv2 and links to the LICENSE file in the Agent QA repository. Read that file to confirm the terms that apply to your use. The source does not state any price.
What do I need to use this skill?
You need the Agent QA MCP server with the run and artifact tools, or a fallback to the dashboard REST APIs or the Agent QA CLI. You also need an Agent QA account or access to the run data the service stores.
Where do I install an agent skill like this?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings, then Capabilities, then Skills.
Does the skill fix my tests or code?
No. The skill classifies the failure and suggests a next step. It explicitly does not modify tests or application code, and it hands code repairs to the separate agent-qa-debug-fix skill.
Why is my triage result low confidence?
Low confidence usually means the run kept few artifacts or logs. The skill lowers confidence when screenshots, DOM or accessibility context, device logs or earlier runs are missing, so retain more evidence to improve results.
Does this skill run scripts or send data over the network?
It does not run scripts. It does use the network, because it fetches run data from Agent QA through the MCP tools or the fallback APIs. The source says it redacts credentials, session tokens and personal data from its report.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run Agent QA tests and want a consistent, evidence-based first look at failed runs, especially when you need to decide who should investigate. Pick something else if you do not use the Agent QA service, if you want an agent to fix code automatically, or if your runs do not keep artifacts and logs, since the classifications will be weak.
Download the skill