Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Triage and analyze Chromium/DevTools LUCI build results and trybot failures, including high-level builder status, compile errors, and detailed test failures. Use this to inspect trybot buildbucket (bb) results for a patchset overview or to query ResultDB directly for comprehensive, actionable information about specific test regressions and infrastructure issues.
.claude/skills/chromedevtools-devtools-ci/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-21 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -17% | 0% |
This skill provides access to specialized triage tools maintained within the depot_tools repository. These tools are used to inspect Buildbucket results and ResultDB test failures.
vpython3 insteadof python3 to ensure all hermetic dependencies (like httplib2) are satisfied.
bb auth-info to proactively check if you arelogged into Buildbucket. If not, inform the user they must run bb auth-login.
The triage scripts and their detailed documentation are located relative to the depot_tools installation:
which gerrit_client.py to find your localdepot_tools root.
bb_wrapper.py): Found in <DEPOT_TOOLS_ROOT>/agents/skills/buildbucket/compile errors, and infra issues.
luci_triage.py): Found in <DEPOT_TOOLS_ROOT>/agents/skills/luci-test-results/ResultDB).
bb log <build_id> <step_name> to see the raw outputfor the failed step.
Once you have located the depot_tools directory, read the following files to understand how to use each tool:
buildbucket: Used for high-level builder status and compile failures.<DEPOT_TOOLS_ROOT>/agents/skills/buildbucket/SKILL.mdluci-test-results: Used for detailed ResultDB test failure analysis.<DEPOT_TOOLS_ROOT>/agents/skills/luci-test-results/SKILL.mdWhen you need to triage a build:
depot_tools as described above.SKILL.md file from depot_tools.bb_wrapper.py or luci_triage.py) using their absolute paths.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | fail→pass | 13,071 | 8,100 | -38% | 1 | 1 | 0% | 2,271 | 2,176 | -4% | 0 | 0 | — |
case-22 | fail→pass | 12,375 | 4,463 | -64% | 1 | 1 | 0% | 2,028 | 1,428 | -30% | 0 | 0 | — |
case-01 | fail→fail | 5,397 | 5,882 | +9% | 1 | 1 | 0% | 865 | 1,014 | +17% | 0 | 0 | — |
case-02 | fail→fail | 4,284 | 4,873 | +14% | 1 | 1 | 0% | 727 | 1,032 | +42% | 0 | 0 | — |
case-03 | fail→fail | 3,992 | 4,606 | +15% | 1 | 1 | 0% | 158 | 808 | +411% | 0 | 0 | — |
case-04 | pass→pass | 7,778 | 6,565 | -16% | 1 | 1 | 0% | 1,492 | 1,972 | +32% | 0 | 0 | — |
case-05 | pass→pass | 6,521 | 5,715 | -12% | 1 | 1 | 0% | 1,352 | 1,614 | +19% | 0 | 0 | — |
case-06 | pass→pass | 7,420 | 3,223 | -57% | 1 | 1 | 0% | 1,190 | 1,118 | -6% | 0 | 0 | — |
case-07 | fail→pass | 10,475 | 4,181 | -60% | 1 | 1 | 0% | 1,927 | 1,461 | -24% | 0 | 0 | — |
case-08 | fail→fail | 12,051 | 7,783 | -35% | 1 | 1 | 0% | 2,528 | 2,242 | -11% | 0 | 0 | — |
case-09 | pass→fail | 4,204 | 1,490 | -65% | 1 | 1 | 0% | 600 | 847 | +41% | 0 | 0 | — |
case-10 | pass→pass | 5,692 | 1,540 | -73% | 1 | 1 | 0% | 905 | 898 | -1% | 0 | 0 | — |
case-11 | fail→fail | 8,143 | 1,987 | -76% | 1 | 1 | 0% | 1,447 | 968 | -33% | 0 | 0 | — |
case-12 | fail→pass | 13,338 | 1,889 | -86% | 1 | 1 | 0% | 2,610 | 971 | -63% | 0 | 0 | — |
case-13 | pass→pass | 8,935 | 1,609 | -82% | 1 | 1 | 0% | 1,495 | 933 | -38% | 0 | 0 | — |
case-14 | fail→pass | 15,125 | 8,590 | -43% | 1 | 1 | 0% | 2,481 | 2,047 | -17% | 0 | 0 | — |
case-15 | fail→pass | 11,378 | 2,640 | -77% | 1 | 1 | 0% | 2,220 | 1,160 | -48% | 0 | 0 | — |
case-16 | fail→pass | 15,824 | 2,924 | -82% | 1 | 1 | 0% | 3,128 | 1,165 | -63% | 0 | 0 | — |
case-17 | fail→pass | 12,805 | 6,115 | -52% | 1 | 1 | 0% | 2,429 | 2,113 | -13% | 0 | 0 | — |
case-18 | pass→pass | 9,176 | 2,767 | -70% | 1 | 1 | 0% | 1,745 | 1,152 | -34% | 0 | 0 | — |
case-19 | fail→pass | 11,201 | 3,685 | -67% | 1 | 1 | 0% | 1,991 | 1,459 | -27% | 0 | 0 | — |
case-20 | pass→pass | 14,944 | 6,361 | -57% | 1 | 1 | 0% | 2,517 | 1,732 | -31% | 0 | 0 | — |
case-23 | fail→fail | 13,925 | 1,736 | -88% | 1 | 1 | 0% | 3,007 | 924 | -69% | 0 | 0 | — |
case-24 | fail→fail | 8,118 | 1,889 | -77% | 1 | 1 | 0% | 1,665 | 990 | -41% | 0 | 0 | — |
case-25 | fail→pass | 5,863 | 1,551 | -74% | 1 | 1 | 0% | 1,026 | 866 | -16% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 23 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.