Repository navigation
Releases
v0.13.0 - deterministic changed-code PR gate
Highlights
gavel pr-check: deterministic commit-snapshot gate for new attributable occurrences on added or replacement lines (exit0/1/2, snippet-free JSON). Historical debt, duplicates, pure renames, and protected base policy stay outside the blocking verdict.- Dead-locator candidates include the declaration line, so a newly added declaration is attributed without blaming an unchanged locator whose consumer was deleted.
- PR snapshots omit unrelated mobile binaries and report assets from the Git object buffer.
- Azure Repos template
templates/azure-pipelines/pr-static-check.ymlcaptures the merge parent, verifies the scanner, and publishes the artifact. Enabling the branch policy remains a separate step.
Install
npm install --save-dev @dsolisp/gavel@0.13.0
npx gavel --versionChanged-code gate
npx --yes @dsolisp/gavel@0.13.0 pr-check <repo> --base <captured-target> --head <captured-merge> --jsonSee CHANGELOG.md and docs/PR_CHECK.md.
v0.12.5 — POM contract beyond C#
Highlights
- POM contract beyond C#:
thin-wrapper,private-locator-alias,new-locator-shadow, andfat-methodnow fire on TypeScript/JavaScript, Python, and Java (same verdicts; no new rule ids). - Precision corpora for
thin-wrapperandprivate-locator-alias(C# + TS). - Hint firewall: do not invent
clickX()wrappers or replace stable ids/By.idwithgetByText(...). new-locator-shadowconfidence: C# medium; non-C# same-file overrides low.
Install
npm install --save-dev @dsolisp/gavel@0.12.5
npx gavel --versionCI gate
npx --yes @dsolisp/gavel@0.12.5 audit --format sarif > gavel.sarifSee CHANGELOG.md.
v0.11.0 — Remediation Loop
v0.11.0 — Remediation Loop
Closes the detect → suggest → fix → verify loop: the scanner now tells agents how to fix findings, cuts false positives on hardcoded-env and brittle-assert, adds adoption/fixture-utilization scanners and CI-history flakiness scoring, and the orchestrator contract mandates post-refactor review and parallel delegation. Also ships GitHub Copilot native discovery and the v0.10.x ecosystem cleanup.
Added
fix:remediation hints inself-check(text/JSON/SARIF), context-aware per waitsubCase; SARIFfixesarray renders inline in Code Scanning.- Adoption scanner (
gavel adoption, report-only):unused-helper+unused-fixture. - Flakiness scoring (
gavel flakiness): Playwright JSON + JUnit/Surefire XML → per-test score; mixed outcome = flaky, fail-every-retry = real failure. CLI-only (not wired intosuite-health.js). - Orchestrator quality gates: post-refactor review + parallel delegation pattern.
- GitHub Copilot native discovery via
.vscode/settings.json(chat.agentSkillsLocations/chat.agentFilesLocations).
Fixed
hardcoded-envenv-wrapper recognition (fallback defaults insideos.environ.get/getenv/process.env).brittle-assertstructure + argument-position parsing (comparison-only; subject-first shapes)..NETbuild dirs (bin/obj/packages/.vs) excluded from scanning.
Tier-R remediation hardening (two adversarial cross-review passes)
brittle-assert: drop trailing failure-message arg for function-style equality APIs (reverses a #11 FP regression);.And.Be(candidate + scan every matcher call so FluentAssertions chains flag.flakiness: self-closing<testcase/>counted correctly; BOM strip; fail-closed on bad input; skipped excluded; order-derivedeventualPass.adoption: comment/string masking (code-only); PlaywrightwaitFor*builtins excluded.
Removed
gavel-omsskill (out of generic scope). Core catalog now 29 skills.
Gate
npm run verify → EXIT=0 — corpus 8/8 @ 100% (brittle-assert TP=19, FP=0, FN=0), diff-corpus 100%, docs/versions/manifest clean, 30/30 unit tests.
Full changelog: v0.10.0...v0.11.0
v0.10.0
gavel v0.10.0 — .NET Ecosystem Parity
Test-quality enforcement now covers the C#/.NET ecosystem end to end, matching the
depth already available for TypeScript, JavaScript, Python, and Java. The same
13-rule constitution — zero new rule tags.
Added
gavel-appium profile. Mobile-native Appium.NET audit: accessibility-first
locators (AppiumBy.AccessibilityId > AndroidUIAutomator / IosNsPredicate >
XPath), native waits, dependency injection, and gesture patterns.
Selenium C#, full audit. Promoted from detect-only to complete constitution
coverage; gavel-selenium gains a C# section (FindElement(By.*),
WebDriverWait / ExpectedConditions, DI, run commands).
Framework detection + freshness. Recognizes Appium.NET (Appium.WebDriver,
pin 8.3.2), Selenium C# (Selenium.WebDriver, pin 4.45.0), and Playwright.NET
(Microsoft.Playwright, pin 1.61.0) from project references, with correct profile
routing. Detection precedence: Appium → Playwright → Selenium.
C# rule coverage. brittle-assert, expect-in-action, selector-leak,
manual-wait, no-di, skip-marker, and hardcoded-env recognize C# idioms
across NUnit, xUnit, MSTest, SpecFlow, Reqnroll, and FluentAssertions — via widened
regexes, no new tags.
Tag extraction. xUnit [Trait] recognized alongside NUnit [Category] and
MSTest [TestCategory]; .cs tag extraction stays runner-agnostic.
dotnet test --filter recipes shared across all three .NET profiles.
Corpus & samples. language: cs corpus at 100% precision, zero false
negatives across every exercising rule. Ships .NET sample repositories
(appium-dotnet, selenium-dotnet, playwright-dotnet) and a SpecFlow/Reqnroll fixture.
Quality gate
npm run verifygreen — 7 version files pinned at 0.10.0; 30 core + 4 companion
skills; 6 parsers; 23/23 unit tests.- Tier-R adversarial cross-review: rounds 1 + 2 remediated.
Deferred (documented non-goals)
- Subject-first
brittle-assert/ prose-literal widening → v0.11.0 (#11). - Additional FluentAssertions/MSTest assertion-regex widenings — dropped: no
corpus-proven, cross-language false-positive-free form. - Runner-name disambiguation for
.cstag extraction — out of scope (would add a
freshness/profile surface).
Interface budget
3 irreversible surfaces (gavel-appium profile, Appium detect/freshness, Selenium C#
routing) — a declared exception to the ≤2 budget, justified as one cohesive .NET
ecosystem release.
Full changelog: v0.9.0...v0.10.0
v0.9.0
V0.9.0 wait intelligence (#15) * feat: Gavel v0.9.0 — Wait Intelligence manual-wait sub-classifier (redundant/stale-read/intentional) with replaceability analysis, threading.Event signal-driven pattern, time-impact estimation, path-weighting, test-vs-utility scope filter, skip-marker prefix suppression, ignore-no-reason context-aware, IDE config directory exclusions, corpus field-level precision locks. Cross-review remediation: closed unset-Event evasion path, audit forwards replaceable/suggestion/pollingLoop, dead code removed, interface budget declared (3 surfaces), migration signal for default excludePaths behavior change. All 7 manifests bumped 0.8.1 → 0.9.0. npm run verify green (19/19). * docs: ship full gavel-refactor agent guidance with v0.9.0 Includes no-op wait migration playbook, violation remediation reference table, cross-step data-flow gate, and selector-leak placement clarification (v0.10.0 items #3, #6, #8 agent-side). CHANGELOG updated with four new entries. * docs * docs
v0.8.1
v0.7.0 — Packaging, Adapters & Onboarding
v0.7.0 — Packaging, Adapters & Onboarding
v0.6.0 — Suite Health Hardening
Full Changelog: v0.6.0...v0.6.0
v0.5.0 — CI Intelligence & Discovery
Full Changelog: https://github.com/dsolisp/gavel/commits/v0.5.0