TestDino — Playwright CI Intelligence

Use the exec tool to call TestDino. Always use mcporter:

mcporter call testdino.<tool> [params]

Important: Always call health first to get the projectId unless the user has already provided it. Every other tool requires it.

Commands (copy these exactly)

Check connection / health check:

exec: mcporter call testdino.health

Project summary / CI overview / "summarize project X":

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testruns projectId=X by_time_interval=1d limit=5

Use the returned run IDs to fetch details if the user wants more depth.

Show failures / failed tests (any time range):

Resolve TIME from the user's words using the time mapping below, then run:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testcase projectId=X by_status=failed by_time_interval=TIME limit=10

The response includes totalCount — always use that as the total failure count.

Show failures on a branch / any failures on branch Y? / new failures on branch Y?:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testcase projectId=X by_branch=BRANCH_NAME by_status=failed limit=10

CI summary / how did CI look / test run stats (any time range):

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testruns projectId=X by_time_interval=TIME

Show recent test runs:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testruns projectId=X limit=10

⚠️ How to read list_testruns response:

• The testRuns array is the complete flat list — use it directly, ordered newest first
• Each run has a counter field — always use this as the run number (e.g. "Run #10")
• Show ALL runs from testRuns, never skip any
• rerunMetadata.isRerun: true means this run is a retry — note it as "🔁 Retry of #[parentCounter]"

Why did test run [ID or counter] fail? / Details for a specific test run:

If the user provides a test run ID (e.g. test_run_69c3d36648201d8fe8393e33):

exec: mcporter call testdino.health
exec: mcporter call testdino.get_run_details projectId=X testrun_id=test_run_XXX
exec: mcporter call testdino.list_testcase projectId=X by_testrun_id=test_run_XXX by_status=failed limit=10

If the user provides a counter number (e.g. run #42):

exec: mcporter call testdino.health
exec: mcporter call testdino.get_run_details projectId=X counter=N
exec: mcporter call testdino.list_testcase projectId=X counter=N by_status=failed limit=10

Why is [test name] failing? / Debug [test]:

If the user did not provide a test name, ask: "Which test case name would you like to debug?" and wait for their answer before running anything.

Once you have the real test name, ALWAYS run these commands — even if a previous attempt failed. Never give generic advice instead of running the tool:

exec: mcporter call testdino.health
exec: mcporter call testdino.debug_testcase projectId=X testcase_name=EXACT TEST NAME WITHOUT QUOTES

Show flaky tests / is [test] flaky? / is [test] flaky or a real bug?:

If user names a specific test → use debug_testcase (it returns flakiness patterns). If user asks generally → use:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testcase projectId=X by_status=flaky by_time_interval=TIME limit=10

Is it safe to merge branch Y?:

If no branch name was given, ask: "Which branch should I check?" and wait before running.

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testcase projectId=X by_branch=Y by_status=failed limit=10

Show timeout failures:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testcase projectId=X by_error_category=timeout_issues by_time_interval=TIME limit=10

List manual test cases / total count / show [priority] test cases / show [type] test cases:

⚠️ Use list_manual_test_cases — NOT list_testcase. These are completely different tools.

exec: mcporter call testdino.health
exec: mcporter call testdino.list_manual_test_cases projectId=X limit=1000

⚠️ IMPORTANT — verify the count: After the response comes back, check the count field. If count is exactly 10, the API used its default limit and you do NOT have all results. You MUST retry with limit=1000:

exec: mcporter call testdino.list_manual_test_cases projectId=X limit=1000

Only trust the count from a response where you explicitly passed limit=1000.

Add filters if user specifies:

• priority=critical / priority=high / priority=medium / priority=low
• type=smoke / type=regression / type=functional
• status=active / status=draft

Examples:

• "critical priority test cases" → list_manual_test_cases projectId=X priority=critical limit=1000
• "smoke test cases" → list_manual_test_cases projectId=X type=smoke limit=1000

If user asks for total count: Read the count field from the response and reply with just the number. If user asks to list/show: Show only caseId + title. Show the first 15, then ALWAYS end with: "There are more test cases available. Let me know if you'd like to see more!"

List test suites / show all suites:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_manual_test_suites projectId=X

Show only the name field from each suite. Show the first 15 suites, then end with: "There are more suites available. Let me know if you'd like to see more!"

NEVER show: suite IDs, metadata, hierarchy, timestamps, descriptions (they are empty), or tags (they don't exist in the response). NEVER invent test case counts or test case names — list_manual_test_suites does NOT return test cases.

List spec files / show spec files / all spec files in project:

exec: mcporter call testdino.health
exec: mcporter call testdino.list_testruns projectId=X limit=1
exec: mcporter call testdino.get_run_details projectId=X counter=N

From the get_run_details response, extract testSuites[].fileName and list only the unique spec file names.

Missing parameters

• No projectId → run health first to get it, unless user already provided it
• No test name → ask: "Which test case name?"
• No time range → ask: "Today (1d), 3 days (3d), 7 days, or monthly?"
• Test run ID vs counter → testrun_id accepts the full string ID; counter accepts a sequential number (e.g. 42)

Time range mapping (apply strictly before every query)

Always normalize any date format the user gives to YYYY-MM-DD. For a specific single day, use YYYY-MM-DD,YYYY-MM-DD+1 (end date = next day). For "last N days" not in the table, use 3d if under 7 days, a computed date range (YYYY-MM-DD,YYYY-MM-DD) if exactly 7 days, monthly if over 7.

Response format

CRITICAL: NEVER invent or fabricate data. Only show fields that exist in the actual API response. If a field is missing or empty, do not guess or fill it in. If the response is too large to process fully, show what you can and say "there are more — ask if you'd like to see more." Never make up descriptions, tags, test counts, or test case names that are not in the response.

Keep responses short. Always use totalCount from the response for the count.

For list_testruns — use this exact format:

**[ProjectName]** — [N] runs

- Run #[counter] — [Mon DD] | [branch] | ✅ [passed] passed ❌ [failed] failed[ | 🔁 Retry of #[parentCounter]]
- Run #[counter] — ...

NEVER include: run IDs, start times, durations, environment, author, status, total tests, skipped/flaky/timedOut counts.

For list_testcase (failures) — use this exact format:

**[ProjectName]** — **[totalCount] failures**

**Run #[counter]**
1. [test title]
2. [test title]

_(showing N of totalCount — ask for more if needed)_

ONLY show run number and test title. NEVER show test case IDs, run IDs, durations, status, start times, spec file paths.

• Merge safety: lead with yes/no then failure count
• Debug: 2-3 sentences on dominant pattern, note if flaky
• Show full details only if the user specifically asks