Skip to main content

​
0.16.0 - 2026-08-30

Scoping release. fix and scan can be pointed at what changed rather than the whole tree, and fix can show what it would do before it writes anything.

​
Added

  • aislop fix --dry-run. Prints the planned fix steps, including skipped steps and why, without writing files. Cannot be combined with an agent handoff or --prompt.
  • aislop fix --changes, --staged, and --base. Restricts fixes to changed or staged files, matching the flags scan already accepted.
  • Changed-line context on findings. Diagnostics are classified as changed-line or existing-file context in terminal, JSON, and SARIF output.

​
Fixed

  • Dependency, audit, and Expo alignment fixes stay inside the selected scope.
  • Scoped fix no longer rewrites test files that a plain fix leaves alone.
  • Scoped fix and scan agree on the score for the same selection.
  • fix respects CI=true as well as CI=1, so spinner sequences do not leak into GitHub Actions logs.
  • C# and C/C++ formatters appear in the dry-run plan.

​
0.15.0 - 2026-08-26

Calibration release. Scores now describe how code is written rather than how much of it there is. Every score moves in this release, most of them upward. Re-baseline CI gates and badges after upgrading.

​
Changed

  • Score reflects finding density, not project size. Deductions are divided by file count before the scoring curve, so a large project and a small one with the same habits land on the same number.
  • Rule caps scale with project size.
  • Default scoring.smoothing lowered from 20 to 5.

​
Fixed

  • ai-slop/trivial-comment no longer flags ordinary terse comments. It now requires a file saturated with them (at least one per 50 non-blank lines).
  • security/vulnerable-dependency reports devDependencies as dev-only at info severity instead of treating them as shipped risk.
  • ai-slop/python-mutable-default no longer flags call-wrapped keyword arguments such as FastAPI Body(default={}).
  • doctor derives languages from the configured scan scope rather than from manifests alone.

​
0.14.1 - 2026-08-08

aislop now supports 10 language targets, with first-class C# and C/C++ analysis and tighter behavior across mixed-language repositories.

​
Added

  • C# support. Project discovery, formatting, linting, complexity analysis, dependency checks, and C#-specific AI-slop rules.
  • C/C++ support. Detection and scoring with cppcheck, clang-tidy, clang-format, complexity analysis, and C/C++-specific AI-slop rules.

​
Fixed

  • Reduced false findings in C# and C++ strings, comments, raw strings, and interpolated code.
  • Improved file discovery, exclusions, invalid-target handling, external-tool timeouts, Windows tool lookup, dependency scans, and Python import checks.

​
Validation

  • 2,032 tests passed across 198 files.
  • Linux and Windows CI passed.
  • The release self-scan scored 100/100 with zero findings.

​
0.14.0 - 2026-07-23

Expanded Python workspace awareness, project discovery, cross-platform diagnostics, and scanner accuracy around real-world source layouts.

​
Changed

  • Resolved dependencies across uv workspaces, including member globs, exclusions, nested packages, editable sources, and shared environments.
  • Made diagnostic paths and external-tool discovery reliable across Windows and Unix-like systems.
  • Preferred project-installed tools over bundled fallbacks and improved bare-source-tree discovery.

​
Fixed

  • Prevented braces in strings, comments, and regular expressions from distorting complexity boundaries.
  • Removed dependency-postinstall coupling and moved npm publication to trusted OIDC publishing.

​
0.13.1 - 2026-06-29

Calibrated hidden-fallback findings, fixed regular-expression comment masking, and refreshed routine dependencies and CI tooling.

​
0.13.0 - 2026-06-28

Added install-channel telemetry, the hidden-fallback rule, framework adapters, more reliable MCP telemetry, and scanner calibration across common false-positive sources.

​
0.12.0 - 2026-06-10

Local agent repair sessions get a full terminal-native workflow, CLI output is tighter across the command surface, and scoring now separates cosmetic cleanup from higher-impact findings.

​
Added

  • aislop agent terminal UI. Foreground agent sessions now render a live terminal UI with steps, activity, edited files, score and target, passes, token usage, and in-terminal decisions. Non-TTY and CI runs keep plain streamed output.
  • Agent workflow commands. Added provider status, connect, default-provider, background sessions, session list/show/watch/apply/stop commands, monitor mode, safe local apply, and optional commit/PR publishing from a reviewed agent diff.
  • Command suggestions. Mistyped commands now get nearest-command guidance instead of falling through to confusing directory-scan errors.
  • Rule impact metadata. JSON output, finding assessment, and aislop rules now expose per-rule score-impact tiers so integrations can distinguish mechanical cleanup from stronger quality and security signals.

​
Changed

  • Scoring calibration. The default ai-slop weight is softer and rule scoring now uses impact multipliers and caps, preserving real findings while reducing score swings from low-impact style signals.
  • Terminal rendering. Scan summaries, badges, rules, doctor, hook, update, trend, provider, and agent outputs use aligned display sections and clearer action guidance.
  • Source filtering and lint context. Generated, vendor, package-manager, and docs-like artifacts are excluded more consistently. ESLint globals and ignores are respected, PHP-served browser JavaScript gets softer no-undef handling, and ruff follows the same scoped file selection as aislop.

​
Fixed

  • Agent telemetry coverage. aislop agent and its provider, plan, monitor, session, apply, watch, and stop subcommands now emit the same lifecycle telemetry as the rest of the CLI, without sending paths, branch names, session ids, prompts, raw findings, or transcripts.
  • Agent subdirectory scope. aislop agent, agent plan, and monitor repair now honor a requested subdirectory in monorepos instead of scanning or fixing the repository root.
  • Agent worktree diffs. Agent apply/publish flows now include staged and untracked files, so provider edits are not skipped when the isolated worktree contains pending changes outside the unstaged diff.
  • Monitor debounce behavior. Monitor mode now waits for stable dirty snapshots without resetting the debounce timer on every poll, while preserving immediate reactions for --debounce 0.
  • Background stop handling. Monitor stop now signals the monitor process group on non-Windows platforms, matching background session stop behavior so active provider subprocesses are not left editing.
  • Installer and CI stability. Tool downloads retry transient GitHub/network failures, the terminal UI dependency stays compatible with the documented Node >=20 floor, and long CLI help-surface tests have explicit CI-safe timeouts.
  • Python and Go calibration. Python manifest/import resolution handles nested manifests and additional stdlib/package aliases, and Go panic heuristics exempt test-support code.

​
0.11.0 - 2026-06-06

PR-scoped quality gates and set-and-forget CI. scan and ci can gate a pull request on only the files it changed with --changes --base <ref>, and an explicit base ref that cannot be resolved fails the run instead of silently passing an empty scan. A moving v1 Action tag plus version: latest mean workflows do not need a version bump, and CI examples standardize on Node 24.

​
Added

  • --base <ref> and ci --changes. Both scan and ci accept --changes --base <ref> to diff against a branch instead of HEAD, so a pull request can be gated on only the files it touches even when those files are already committed.
  • forceFixable flag. scan --json diagnostics carry an explicit forceFixable boolean for findings that only aislop fix -f resolves.

​
Changed

  • Set-and-forget CI. A moving v1 Action tag tracks the latest release, so workflows can pin uses: scanaislop/aislop@v1 and keep version: latest.
  • Node standardized on 24. CI examples use Node 24. The supported runtime floor remains Node >=20.

​
0.10.2 - 2026-06-02

A patch release focused on safer release and CI plumbing, sharper scan consistency, and rule catalog alignment.

​
Added

  • Inline suppression. Suppress known intentional findings with aislop-ignore-next-line, aislop-ignore-line, or aislop-ignore-file, optionally scoped to specific rule IDs and a reason.
  • .aislopignore. Keep whole paths out of every scan with a root-level ignore file.
  • aislop fix --safe. Restrict auto-fix to reversible fixes such as unused imports, import merging, narrative-comment removal, and formatting.

​
Fixed

  • Exact-version Action pins. Exact npm pins run from an isolated temp directory so npm cannot resolve checked-out local package metadata instead of the published package.
  • Rule catalog consistency. aislop rules, rule labels, and docs cover implemented rule IDs consistently.