메뉴
Measure and improve Provenant's Rust test coverage locally with cargo-llvm-cov. Use for code coverage, llvm-cov, finding untested code, test gaps, or "what isn't tested".
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Maintain Provenant generated docs and their xtask checks. Use for supported formats, output field reference, serve OpenAPI, benchmark chart, generated docs drift, validate-urls, or docs check failures.
Iterate on compare-outputs for a concrete repository or artifact until only justified Provenant advantages remain, then record benchmark-backed results and open a PR.
Triage Provenant CI failures and map GitHub jobs to local reproduction commands, owning docs, and existing skills. Use for CI failed, check.yml, clippy, cargo deny, unused deps, license headers, generated docs drift, golden shard, or workflow failure.
Open a Provenant PR with the repo template and drive it to merge-ready — wait for CI and Greptile/human review, address comments by amending + force-pushing, resolve review threads, and iterate until zero unresolved. Use for opening a PR, handling review/Greptile comments, resolving review threads, or PR CI signals.
Measure whether a Provenant code change actually improves performance — profile-first, before/after on the same code, regression-guarded, with a byte-identical correctness gate. Use for "is this optimization actually faster", "benchmark my change", "profile a copyright/license hotspot", "before/after perf", "did my refactor regress performance", or "prove this speedup".
Implement a new package parser in Provenant, following the full workflow from research through registration, testing, assembly wiring, and validation.
| name | code-coverage |
| description | Measure and improve Provenant's Rust test coverage locally with cargo-llvm-cov. Use for code coverage, llvm-cov, finding untested code, test gaps, or "what isn't tested". |
Coverage in Provenant is a local tool for finding untested code, not a CI gate
or a number to maximize. See docs/TESTING_STRATEGY.md
for the "confidence, not coverage" philosophy. Use this to locate genuinely
untested paths, then write meaningful behavior tests for them — do not add tests
just to move the number.
cargo-llvm-cov is installed by npm run setup. Otherwise:
cargo install --locked cargo-llvm-cov
# Per-file summary table for the whole library (fast triage)
npm run coverage # = cargo llvm-cov --lib
# Annotated line-by-line HTML report, opened in a browser
npm run coverage:html # = cargo llvm-cov --lib --open
# Scope to one module while iterating (much faster)
# The test-name filter must follow `--` so it is forwarded to the test binary.
cargo llvm-cov --lib -- copyright::refiner
# Show uncovered regions for a path as text
cargo llvm-cov --lib --show-missing-lines -- <filter>
--lib covers unit tests only. Integration suites (tests/*.rs) and
golden suites (--features golden-tests) exercise far more code than this
credits, so treat the number as a lower bound. A file showing 0% may still
be well-covered by golden/integration tests.copyright/grammar/rules.rs) or generated
code is expected and not worth "fixing".npm run coverage and pick a module with thin coverage on real logic.cargo llvm-cov --lib --show-missing-lines -- <module>) to see the
exact uncovered lines.tests/ and *_scan_test.rs).docs/TESTING_STRATEGY.md
(right layer: unit vs scan/assembly contract vs golden).