slop

A code quality linter for codebases where AI agents are writing most of the diffs.

Static-analysis defaults were calibrated for codebases where a productive human wrote ~100 lines on a busy day and another human reviewed every one. An agent can drop that much into a single file before emitting its first status message, and the damage — deep coupling, WMC-heavy classes, duplicated handlers, files that grow 500 LOC in a week, identifier soup that bloats every downstream prompt — lands inside one session rather than over quarters. slop runs well-cited metrics across three substrates — structural (McCabe, Chidamber & Kemerer, Nejmeh, Martin, Tornhill, Campbell), information-theoretic (Halstead), and lexical (identifier vocabulary) — at thresholds tuned for that pace.

Example

$ slop lint

slop 1.0.0 — scanning .

structural.complexity
  cyclomatic
    ✗ src/pipeline/ingest.py:44 process_batch — CCX 18 exceeds 10
    ✗ src/pipeline/ingest.py:112 _normalize_rows — CCX 14 exceeds 10
  cognitive
    ✗ src/pipeline/ingest.py:44 process_batch — CogC 26 exceeds 15
  3 violations, 142 checked

structural.hotspots (14 days ago, 87 commits)
  ✗ src/pipeline/transformation.py — CCX=45, growth +367 LOC
  ✗ src/lifecycle/tasks/write.py — CCX=41, growth +556 LOC
  2 violations

structural.duplication
  ⚠ . — clone density 7.2% exceeds threshold 5.0% (12 cloned functions across 4 clusters)
  ⚠ src/api/handlers/users.py:88 update_user — function 'update_user' is a Type-2 clone (fingerprint a3b1c4d2f001) — also at: src/api/handlers/orgs.py:74, src/api/handlers/teams.py:81
  2 violations

information.volume
  ✗ src/pipeline/ingest.py:44 process_batch — Volume 2147 exceeds 1500
  1 violation, 412 checked

information.magic_literals
  ⚠ src/billing/discount.py:31 apply_promo — 'apply_promo' contains 5 distinct magic numeric literals (threshold: 3): 7, 14, 30, 86400, 0.15
  1 violation

lexical.stutter
  ⚠ src/services/user_service.py:47 user_update_user_profile — identifier 'user_update_user_profile' repeats tokens ['user'] from enclosing class 'UserService'
  ⚠ src/services/user_service.py:91 get_user_user_id — identifier 'get_user_user_id' repeats tokens ['user'] from enclosing class 'UserService'
  2 violations, 412 checked

────────────────────────────────────────
6 violations | 5 advisories | 22 rules checked | FAIL

Exit 0 clean, 1 on violations, 2 on error. Works in CI, pre-commit, and interactively.

Install

pip install agent-slop-lint

slop shells out to rg, fd, and git. Install via your system package manager (apt install ripgrep fd-find git, brew install ripgrep fd git, or equivalent) and run slop doctor to verify. Full per-platform steps, CI recipes, and pre-commit wiring are in the setup guide.

Rules

slop ships 25 rules across three suites:

• structural.* — control-flow complexity, CK class metrics, hotspots, package distance, dependency cycles, duplication, god modules, type-discipline rules.
• information.* — Halstead volume and difficulty, magic literals, section-divider comments.
• lexical.* — identifier verbosity, tersity, and stutter against the enclosing scope.

The full rule index with default thresholds, citations, and per-rule pages lives in docs/rules/. For threshold tuning and the default / lax / strict profiles, see the configuration reference.

Languages

Language-specific caveats (JavaScript packages, Rust deps, Julia CK metrics, C class metrics, C/C++ -I-path resolution and out-of-line method attribution, Ruby type-discipline rules silent-skip and open-class WMC aggregation) are documented in the C notes, the C++ notes, the Julia notes, the Ruby notes, and the relevant rule pages.

CLI

slop lint                         Run all enabled rules
slop check <category|rule>        Run one category or rule
slop init [default|lax|strict]    Generate .slop.toml
slop doctor                       Check fd, rg, git are installed
slop hook                         Install a git pre-commit hook
slop skill <dir>                  Install the bundled agent skill
slop rules                        List rules with thresholds
slop schema                       Config schema as JSON

Output formats are --output human (default), --output json (CI and agents), and --output quiet. Run slop --help for the full flag list.

Configuration

slop walks upward from CWD looking for .slop.toml first, then pyproject.toml with a [tool.slop] table — the same discovery ruff and mypy use. root resolves relative to the config file's directory. --config and --root on the CLI override both.

slop init          # balanced defaults
slop init lax      # legacy or gradual adoption
slop init strict   # greenfield or quality-focused

Every threshold, profile, and waiver mechanism is documented in the configuration reference.

Architecture

slop ships its own discovery primitives and metric kernels. Each rule is a thin wrapper that loads config, calls a deterministic kernel, and emits Violation objects for threshold breaches. Primitives are organised by substrate — _fs (fd), _text (ripgrep), _ast (tree-sitter) — with cross-tool primitives in _compose and metric kernels in _structural and _lexical. One pip install gives you the whole thing; no companion runtime.

Acknowledgments

slop implements metrics from McCabe, Halstead, Chidamber & Kemerer, Nejmeh, Martin, Lakos, Tornhill, and Campbell. Full bibliography in NOTICE. AI assistance and contributor credits in the project CITATIONS.

License

Apache 2.0. See LICENSE.