safere

Prepare a daily SafeRE maintainer decision-support report over trusted GitHub PRs and issues: summarize what changed since the last review, run per-PR review/benchmark work only when needed, assess linked issue/PR state, and write durable Markdown reports and artifacts without posting or pushing.

Run a serialized background sweep of open SafeRE GitHub PRs: skip drafts, track reviewed PR head SHAs and discussion changes, assess PR intent against implementation, run the review-fix-loop skill for P2+ findings, reproduce optimization benchmark claims against current main, and write durable reports and artifacts without pushing or commenting.

Run Codex's noninteractive code review on a Git working tree, fix review findings at or above a requested priority such as P2, run local verification, and repeat until no in-scope findings remain. Use when the user asks to run /review, codex exec review, or a review/fix loop over branch changes against main, uncommitted changes, or a specific commit.

Review SafeRE/java.util.regex divergence sweep outputs, JSONL reports, fuzz result sets, or differential matrices by bucketing related failures, classifying whether each bucket is specification-required, reasonable compatibility behavior, intentional JDK implementation detail, or unsupported under SafeRE's linear-time guarantee, and assessing whether each bucket can be fixed with a principled linear-time implementation. Use when asked to analyze, inventory, categorize, triage, summarize, or decide next steps for divergence results rather than immediately fixing a single bug.

Review SafeRE source files to assess whether the linear-time regex matching guarantee still holds. Use when asked to audit the whole project, a module, or a change for risks such as backtracking, repeated rescans, unbounded recursion, cache-state blowups, or parser acceptance of non-linear regex features.

Fix or classify SafeRE bugs where SafeRE behavior diverges from java.util.regex using the project bug-fixing discipline: JDK 26 Pattern/Matcher Javadoc assessment, test-driven diagnosis for behavior changes, principled class-level fixes instead of point patches, invariant-based verification, and regression/fuzz coverage where applicable. Use when a SafeRE bug involves any difference from the JDK in regex syntax parsing, parse-time acceptance or rejection, matching results, capture semantics, quantified captures, group boundaries, find() sequences, regions/bounds, hitEnd()/requireEnd(), matcher state transitions, replacement behavior, or crashes/errors from Parser, Pattern.compile, Pattern, or Matcher.

Triage SafeRE Jazzer fuzzer runs performed manually by the user: locate Surefire/Jazzer logs, distinguish latest run artifacts from persistent corpus and older saved inputs, extract AssertionError findings, summarize self-contained repro details, and prepare GitHub issue material without rerunning fuzzing.