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".
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".
Implement a new package parser in Provenant, following the full workflow from research through registration, testing, assembly wiring, and validation.
Maintain Provenant's Rust dependency policy and Cargo dependency changes. Use for cargo add/remove/update, Cargo.toml, Cargo.lock, deny.toml, cargo-deny, cargo-machete, unused deps, dependency licenses, advisories, or crate-size policy.