Imported from vasceannie/slopgate (bundle/shared/skills/slopgate-hygiene-orchestrator/SKILL.md). Install upstream with npx skills add vasceannie/slopgate --skill slopgate-hygiene-orchestrator. Copyright stays with the author (MIT).

Hygiene Orchestrator

Coordinate systematic elimination of linter warnings/errors using parallelized agents, JSON-exported findings, and persistent progress tracking.

When to Use

Use when the task is repo-wide, multi-file, or multi-rule Slopgate cleanup.

• User asks to fix all warnings, clean up lints, pass the quality gate, or coordinate hygiene work.
• slopgate lint check --details produces many findings that need grouping, tracking, and dispatch.
• Multiple files or collectors are involved and progress needs a persistent .hygiene/ tracking artifact.
• Use this as the coordinator; delegate local structural repairs to slopgate-code-hygiene-refactor and test gaps to slopgate-test-extender.

When Not to Use

Do not use for a single hook denial or one touched file. Use slopgate-code-hygiene-refactor for local structure, slopgate-code-smell-utility-locator for helper/constant ownership, and slopgate-test-extender for test-only findings.

Do not use when the task is only to inspect current telemetry or explain a rule without applying a cleanup plan.

Critical Rules

1. NEVER run linters directly in terminal — Use slopgate lint check (repo root, no path args) as the primary gate. Discover project-specific tools by inspecting package.json, pyproject.toml, Cargo.toml, etc.
2. Always export to JSON — Lint outputs must be parsed into structured JSON for tracking and dispatch
3. Update tracking before context compaction — Progress must survive context limits via .hygiene/tracking.json
4. Remind all agents about related skills — Invoke /type-strictness for type fixes, /code-hygiene-refactor for structural issues, /test-extender for coverage gaps

Workflow Overview

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│  1. Export      │───>│  2. Parse &      │───>│  3. Dispatch    │
│  Lint Results   │    │  Group Issues    │    │  Parallel Agents│
└─────────────────┘    └──────────────────┘    └─────────────────┘
         ^                                              │
         │              ┌──────────────────┐            │
         └──────────────│  4. Validate &   │<───────────┘
                        │  Update Tracking │
                        └──────────────────┘

Phase 1: Export Lint Results

Primary: slopgate lint check

Run from the repo root. No path argument is accepted.

# Basic run
HOME=/home/trav /home/trav/.local/bin/slopgate lint check

# Detailed output with repair guidance
HOME=/home/trav /home/trav/.local/bin/slopgate lint check --details

# Export to JSON for parsing (pipe stdout to a parser)
HOME=/home/trav /home/trav/.local/bin/slopgate lint check --details 2>&1 | \
  python3 /path/to/skill/scripts/parse_slopgate_lint.py --output .hygiene/slopgate_lint.json

The --details flag adds per-finding fields: file, location, signature, detail, stable-id, prognosis, scaffold, metadata.*, agent-context, validation, verify.

Secondary: project-specific lint tools

If the project uses additional linters beyond Slopgate, discover the correct invocation rather than assuming a Makefile:

# Look for common entrypoints
ls package.json 2>/dev/null && npm run lint --silent 2>/dev/null
ls pyproject.toml 2>/dev/null && python3 -m ruff check . 2>/dev/null
ls Cargo.toml 2>/dev/null && cargo clippy --all-targets 2>/dev/null

Pipe any secondary lint output through a parser into .hygiene/ JSON or text. Do not assume make targets exist.

Phase 2: Parse and Group Issues

Parsing Slopgate Lint Output

Use scripts/parse_slopgate_lint.py to parse --details output into structured JSON:

python /path/to/skill/scripts/parse_slopgate_lint.py \
  --input <(slopgate lint check --details 2>&1) \
  --output .hygiene/slopgate_lint.json

The parser produces:

{
  "header": {
    "project": "/path/to/repo",
    "baseline": "/path/to/repo/baselines.json"
  },
  "summary": {
    "collectors": [
      {"name": "long-method", "total": 3, "new": 2},
      {"name": "untested-production-code", "total": 1, "new": 1}
    ],
    "total_findings": 4,
    "new_findings": 3,
    "baseline_findings": 1,
    "fixed_findings": 0
  },
  "findings": [
    {
      "status": "NEW",
      "collector": "long-method",
      "fields": {
        "file": "src/main.py",
        "location": "src/main.py:52",
        "signature": "callable `f`",
        "detail": "lines=52",
        "stable-id": "long-method|src/main.py|f|lines=52",
        "prognosis": "function is doing multiple phases inline.",
        "scaffold": "extract named helpers for parse/validate/transform/persist/render phases; keep data flow explicit."
      }
    }
  ]
}

Grouping Strategy

Group issues into non-conflicting batches for parallel agent work:

Non-conflicting batch rules:

• Files in batch A share NO imports with files in batch B
• Type definition files before files that use those types
• Base classes before derived classes

Use stable-id to deduplicate findings across iterations.

Phase 3: Dispatch Parallel Agents

Agent Configuration

Each agent batch receives:

1. File list: Non-overlapping with other batches
2. Issue manifest: Specific findings with stable-id, scaffold, and location
3. Context reminder: Relevant skills, project conventions

Dispatch Template

# For each batch, spawn code-hygiene-enforcer agent:
batch_1:
  files: [src/noteflow/grpc/_mixins/annotation.py, src/noteflow/grpc/_mixins/calendar.py]
  findings:
    - stable-id: "long-method|src/noteflow/grpc/_mixins/annotation.py|parse|lines=65"
      scaffold: "extract named helpers for parse/validate/transform phases"
  agent: code-hygiene-enforcer
  instructions: |
    Fix the following issues. Remember: use /type-strictness skill for all type fixes.
    Never add # type: ignore. Update all dependent code.

batch_2:
  files: [src/noteflow/grpc/_mixins/project/_membership.py, src/noteflow/grpc/_mixins/project/_mixin.py]
  findings:
    - stable-id: "untested-production-code|src/noteflow/grpc/_mixins/project/_membership.py|coverage-001"
      scaffold: "add behavior tests for the unreferenced public symbols"
  agent: code-hygiene-enforcer
  instructions: |
    Fix None-safety issues. Add proper guards or update function signatures.
    Use /test-extender for test additions. Use /type-strictness for type refinements.

Parallel Execution

Use Task tool to launch multiple agents simultaneously:

Task(subagent_type="code-hygiene-enforcer", prompt="...", run_in_background=true)
Task(subagent_type="code-hygiene-enforcer", prompt="...", run_in_background=true)
...

Monitor with TaskOutput(task_id=..., block=true).

Phase 4: Validate and Update Tracking

Post-Agent Validation

After each agent completes:

1. Re-run slopgate lint:

   HOME=/home/trav /home/trav/.local/bin/slopgate lint check --details 2>&1 | \
     python3 /path/to/skill/scripts/parse_slopgate_lint.py --output .hygiene/slopgate_lint.json

2. Verify no regressions with the project's test command:

   # Discover the test runner; don't assume make
   ls pytest.ini pyproject.toml tox.ini 2>/dev/null && python3 -m pytest -q
   ls package.json 2>/dev/null && npm test --silent 2>/dev/null
   ls Cargo.toml 2>/dev/null && cargo test --quiet 2>/dev/null

3. Check for new issues introduced by fixes

Update Tracking File

Maintain .hygiene/tracking.json:

{
  "session_id": "2025-01-02T03:45:00Z",
  "iteration": 3,
  "repo": "/path/to/repo",
  "baseline": "/path/to/repo/baselines.json",
  "initial_counts": {
    "long-method": 5,
    "untested-production-code": 3,
    "repeated-code-block": 12
  },
  "current_counts": {
    "long-method": 1,
    "untested-production-code": 0,
    "repeated-code-block": 4
  },
  "completed_items": [
    {
      "stable_id": "long-method|src/foo.py|process|lines=68",
      "file": "src/foo.py",
      "collector": "long-method",
      "resolved_at": "2025-01-02T04:00:00Z"
    }
  ],
  "pending_items": [
    {
      "stable_id": "long-method|src/bar.py|transform|lines=72",
      "file": "src/bar.py",
      "collector": "long-method",
      "scaffold": "extract named helpers for transform/persist phases",
      "status": "open"
    }
  ],
  "blocked_items": [
    {
      "stable_id": "repeated-code-block|src/qux.py|helper|lines 10-15",
      "file": "src/qux.py",
      "collector": "repeated-code-block",
      "reason": "Requires architectural change — shared helper needs design review"
    }
  ],
  "last_updated": "2025-01-02T04:15:00Z"
}

Use stable-id as the primary key for tracking individual findings across iterations.

Context Compaction Protocol

CRITICAL: Before context limit is reached, MUST update:

1. tracking.json with current progress
2. Todo list with remaining work items
3. Memory (if using Serena): Write to hygiene_progress memory

Handoff Message Template

hygiene_handoff:
  status: "in_progress"
  iteration: 4
  remaining_errors: 12
  next_batch:
    files: [list of files]
    findings:
      - stable-id: "collector|file|signature|detail"
        scaffold: "repair guidance from slopgate output"
  instructions: |
    Continue hygiene orchestration from iteration 4.
    Read .hygiene/tracking.json for full state.
    Run /hygiene-orchestrator to resume.

Iteration Loop

WHILE issues_remaining > 0:
    1. Export fresh lint results (slopgate lint check --details)
    2. Parse into JSON (parse_slopgate_lint.py)
    3. Group into non-conflicting batches by stable-id
    4. Dispatch agents in parallel
    5. Wait for completion
    6. Validate fixes (re-run slopgate lint check)
    7. Update tracking.json
    8. IF context_nearing_limit:
           Write handoff state
           EXIT with resume instructions
    9. Increment iteration

Available Commands

Project-specific tools (discover, don't assume):

Skill Reminders

Always remind dispatched agents:

1. Use /type-strictness skill for type annotations
2. Use /test-extender skill when modifying tested code or adding coverage
3. Use /code-hygiene-refactor skill for structural issues (flat siblings, oversized modules, god classes)
4. Never use # type: ignore — fix the underlying issue
5. Never use Any unless absolutely unavoidable
6. Update all callers when changing signatures
7. Run tests after fixes: pytest <affected_tests>

Error Categories Reference

Slopgate collectors (common)

• long-method — Function exceeds line threshold; extract helpers
• repeated-code-block — Duplicated behavior; extract shared helper/service
• untested-production-code — Public symbols lack test references; add behavior tests
• flat-sibling-modules — Files share prefix; convert to sub-package (see /code-hygiene-refactor)
• oversized-module — Module exceeds line threshold; split by responsibility
• god-class — Class has too many methods; extract collaborators

Python (pyrefly)

• unbound-name — Variable may be uninitialized
• missing-attribute — Method called on None
• bad-argument-type — Type mismatch in function call
• untyped-import — Missing type stubs
• not-iterable — Async iteration issues

TypeScript (biome/tsc)

• noExplicitAny — Replace Any with specific type
• noUnusedVariables — Remove or use the variable
• useConst — Prefer const over let

Rust (clippy)

• clippy::unwrap_used — Handle errors properly
• clippy::expect_used — Provide context or use ?
• dead_code — Remove unused code

Resumption Protocol

If starting fresh after compaction:

1. Read .hygiene/tracking.json
2. Read .hygiene/slopgate_lint.json
3. Continue from tracking.json.iteration
4. Process tracking.json.pending_items by stable-id