Inspects LUCI build details using the inspect_coverage_build.py script to validate code coverage configurations, test execution, and target file coverage.
76%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
—
The risk profile of this skill
Fix and improve this skill with Tessl
tessl review fix ./agents/skills/experimental-code-coverage-build-inspector/SKILL.mdThis skill inspects LUCI builds to validate instrumentation configuration, coverage pipeline execution, recipe skip/error messages, and target file coverage metrics.
[!IMPORTANT] Mandatory Diagnostic Script: You MUST use the automated inspection script
tools/code_coverage/inspect_coverage_build.pyas your primary inspection tool. When investigating a specific error flagged by the script, use thebuildbucketskill located indepot_toolsrather than manually runningbb get.
gn_args and .gclient configuration for a LUCI tryjob.overall test coverage and unit test coverage pipeline steps.When invoked during CQ triage, this skill requires:
build_link: The Buildbucket build URL/link to inspect (e.g.,
https://ci.chromium.org/b/8679...), passed in the prompt by calling agent.language_context: Programming language context ("cpp", "objc", "rust",
"java", "js", or "ts"). Consult references/language_coverage_map.json.target_files: Optional list of specific target files under investigation
(e.g., ["chrome/browser/ui/color/material_new_tab_page_color_mixer.cc"]).test_suites: Specific list of target test suites (e.g.,
["chrome_junit_tests", "content_browsertests"]) to verify execution for.metrics_of_concern: Specific coverage metric types or target thresholds
under investigation (e.g., "line_coverage", "function_coverage").Execute the automated inspection tool against the target build:
python3 tools/code_coverage/inspect_coverage_build.py \
--build <build_link> \
--language <language_context> \
--files <target_file_1> <target_file_2> \
--save-artifact scratch/inspection_<build_id>.mdSCHEDULED or
STARTED), do NOT exit or return a final report prematurely.schedule tool to set up a recurring 10-minute poll:
CronExpression: "*/10 * * * *"Prompt: "Re-run tools/code_coverage/inspect_coverage_build.py against
<build_link>. If still running, go idle. If terminal, cancel this cron
timer using manage_task and execute Section 3 (Conditional Deep-Dive)."Once the script completes against a terminal build, use the buildbucket skill
to pull deeper log snippets only when specific conditions occur:
SUCCESS and target files have expected coverage, no
extra log scraping is needed. Proceed directly to Section 4.SUCCESS:
inspect_coverage_build.py extracts 0.00% Line Coverage (or very low
coverage) for target_files despite pipeline SUCCESS:
buildbucket to check which test suites actually ran on the bot.GN args or .gclient):
buildbucket to inspect the lookup GN args step output.Compile a structured Markdown artifact (scratch/inspection_<build_id>.md or
path requested by parent agent) structured into three parts:
Report back to the parent CQ Debugging agent with the artifact file URI.
The inspect_coverage_build.py script automatically verifies three phases:
references/language_coverage_map.json.llvm-cov HTML reports
from GCS and parses exact coverage percentages and uncovered line numbers.[!IMPORTANT] Use the
buildbucketSkill: For all follow-up log investigation, step status queries, and error log extraction listed below, you MUST use thebuildbucketskill located indepot_tools. If thebuildbucketskill is not available indepot_tools, notify the user.
When the script detects official Clang coverage recipe skip or error messages
from build/recipes/recipe_modules/code_coverage/api.py, follow these triage
instructions to investigate further:
skip processing clang coverage data because no profile data collected
.profraw profile dirs.use_clang_coverage = true was missing.skip processing because no profdata was generated
llvm-profdata merge failed or produced no output.Found invalid profraw files (presentation.properties['merge errors'])
.profraw files detected during profile merge.merge errors property output to identify crashing or
OOM-killed Swarming test shards.skip processing because no data is found
target_files ran
on this bot or if target_files lack unit tests.process_coverage_data_failure = True
processing_step.logs['error'] for Python traceback.chromium/chromium
bd7264f
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.