Agent skill

explore

Navigate and understand the MenGrowth codebase (both curation and preprocessing pipelines). Use when asked to find where something is implemented, understand how a phase/step works, trace data flow, debug a rejection or preprocessing failure, or identify where to add new functionality.

View SKILL.md on GitHub Repository
Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/explore-mariopasc-mengrowth

SKILL.md

Explore — MenGrowth Codebase Navigation

When to Use

  • "Where is X implemented?"
  • "How does phase/step N work?"
  • "Why was patient/file X rejected?"
  • "Where would I add a new quality check / modality / plot / preprocessing step?"
  • "What calls what?" or "Trace the data flow for ..."
  • Any task that requires understanding the codebase before modifying it

Quick Module Map

mengrowth/
  cli/
    curate_dataset.py              ← Curation pipeline orchestrator
    preprocess.py                  ← Preprocessing pipeline CLI
  preprocessing/
    config.py                      ← Curation config dataclasses (grep for @dataclass)
    utils/
      reorganize_raw_data.py       ← Phase 1 (scan_source_baseline, scan_source_controls, copy_and_organize)
      filter_raw_data.py           ← Phase 2 (filter_dataset) + Phase 4 (reid_patients_and_studies)
      quality_filtering.py         ← Phase 3 (run_quality_filtering, validate_file, _validate_patient)
      metadata.py                  ← MetadataManager, PatientMetadata, Excel parsing
    quality_analysis/
      analyzer.py                  ← Phase 5 (QualityAnalyzer.run, scan_dataset, analyze_patient)
      metrics.py                   ← Per-image metric functions (SimpleITK-based)
      visualize.py                 ← Phase 6 (QualityVisualizer, generate_html_report)
    src/
      config.py                    ← Preprocessing config (StepRegistry, StepMetadata, all step configs)
      preprocess.py                ← PreprocessingOrchestrator (step execution engine)
      base.py                      ← BasePreprocessingStep ABC (execute + visualize)
      steps/
        step_registry.py           ← Re-exports StepRegistry from config
        data_harmonization.py      ← Step 1 handler
        bias_field_correction.py   ← Step 2 handler
        resampling.py              ← Step 3 handler
        cubic_padding.py           ← Step 4 handler
        registration.py            ← Step 5 handler
        skull_stripping.py         ← Step 6 handler
        intensity_normalization.py ← Step 7 handler
        longitudinal_registration.py ← Step 8 handler
      data_harmonization/          ← NRRD→NIfTI (io.py), reorientation (orient.py), head masking
      bias_field_correction/       ← N4 via SimpleITK (n4_sitk.py)
      resampling/                  ← BSpline, ECLARE, Composite
      registration/                ← ANTs registration (nipype + antspyx implementations)
      skull_stripping/             ← HD-BET (hdbet.py), SynthStrip (synthstrip.py)
      normalization/               ← Z-score, KDE, FCM, WhiteStripe, PercentileMinMax, LSQ
      checkpoint.py                ← Checkpoint/resume support

Exploration Strategies

Strategy 1: "Where is feature X?"

Looking for... Go to
A CLI flag or argument (curation) cli/curate_dataset.py → argparse section
A CLI flag or argument (preprocessing) cli/preprocess.py → argparse section
A curation threshold or default preprocessing/config.py → search the dataclass name
A preprocessing config or default preprocessing/src/config.py → search the dataclass name
A quality check (A1–E1) preprocessing/utils/quality_filtering.py → search by check ID or function
Modality synonym mapping preprocessing/config.py → RawDataConfig.standardize_modality()
Patient ID normalization preprocessing/utils/reorganize_raw_data.py → extract_patient_id()
Re-identification logic preprocessing/utils/filter_raw_data.py → reid_patients_and_studies()
A curation plot or visualization preprocessing/quality_analysis/visualize.py
Clinical metadata field preprocessing/utils/metadata.py → PatientMetadata dataclass
Rejection tracking Search for rejected_files or rejection_reason across utils/
Step execution level preprocessing/src/config.py → STEP_METADATA dict
Step handler registration preprocessing/src/preprocess.py → _register_step_handlers()
Brain mask resolution Step handler's _resolve_mask_path() or search for brain_mask
Registration diagnostics preprocessing/src/registration/diagnostic_parser.py
Normalization methods preprocessing/src/normalization/ → one file per method

Strategy 2: "How does data flow through phase/step N?"

Curation phases:

Phase Entry Function Key Subfunctions
1 reorganize_raw_data() scan_source_baseline(), scan_source_controls(), copy_and_organize()
2 filter_dataset() normalize_study_sequences(), remove_non_required_sequences()
3 run_quality_filtering() _validate_patient() → validate_file(), study-level, patient-level
4 reid_patients_and_studies() Inline in filter_raw_data.py
5 QualityAnalyzer.run() scan_dataset(), analyze_patient(), compute_*_metrics()
6 QualityVisualizer.generate_all() Individual plot methods, generate_html_report()

Preprocessing steps:

Step Handler Location Key Implementation
1 steps/data_harmonization.py data_harmonization/io.py, orient.py, head_masking/
2 steps/bias_field_correction.py bias_field_correction/n4_sitk.py
3 steps/resampling.py resampling/bspline.py, eclare.py, composite.py
4 steps/cubic_padding.py Inline padding logic
5 steps/registration.py registration/multi_modal_coregistration.py, intra_study_to_atlas.py
6 steps/skull_stripping.py skull_stripping/hdbet.py, synthstrip.py
7 steps/intensity_normalization.py normalization/zscore.py, kde.py, etc.
8 steps/longitudinal_registration.py registration/longitudinal_registration.py

Strategy 3: "Why was patient/file X rejected?"

  1. Open quality/rejected_files.csv
  2. Filter by patient_id or filename
  3. Check stage column:
    • 0 → Phase 1 (reorganization): file excluded by glob pattern, unrecognized modality, or duplicate
    • 1 → Phase 2 (completeness): missing sequences or insufficient studies
    • 2 → Phase 3 (quality): failed a blocking check
  4. If stage=2, cross-reference with quality/quality_issues.csv for the specific check_name
  5. To find the check implementation: search quality_filtering.py for the check_name value

Strategy 4: "Where do I add a new ___?"

Adding... Steps
Quality check 1. Add config @dataclass in config.py under QualityFilteringConfig. 2. Add function in quality_filtering.py returning ValidationResult. 3. Wire into validate_file() or study/patient level.
Preprocessing step 1. Create handler in src/steps/. 2. Add StepMetadata entry in src/config.py. 3. Add config @dataclass in src/config.py. 4. Register handler in preprocess.py → _register_step_handlers().
Modality Add synonyms in configs/raw_data.yaml → modality_synonyms. If required, add to FilteringConfig.sequences.
Source directory Add scanner function in reorganize_raw_data.py. Call from reorganize_raw_data().
Plot Add method to QualityVisualizer. Add toggle in PlotConfig. Call from generate_all().
Metric Add function in metrics.py. Wire into analyze_patient() in analyzer.py.
Normalization method Add class in src/normalization/. Import in preprocess.py. Add to config dispatch.
Registration variant Add implementation in src/registration/. Update factory in factory.py.

Key Patterns to Recognize

Configuration Pattern

Every tunable parameter is a field in a @dataclass with a default value, parsed from YAML:

python
@dataclass
class SomeCheckConfig:
    enabled: bool = True
    threshold: float = 5.0
    action: str = "block"

Validation Result Pattern (Curation)

Every quality check returns:

python
ValidationResult(passed=bool, check_name=str, message=str, action="warn"|"block", details=dict)

Step Execution Pattern (Preprocessing)

python
StepExecutionContext(patient_id, study_dir, modality, paths, orchestrator, step_name, step_config)

Handlers receive context and dispatch to implementation classes (e.g., ZScoreNormalizer, BSplineResampler).

Parallel Execution Pattern

  • CPU-bound → ProcessPoolExecutor at patient granularity
  • I/O-bound → ThreadPoolExecutor at file granularity
  • Results sorted by ID after collection for determinism
  • All arguments must be picklable (pure dataclasses, Paths, primitives)

Temp-File Pattern (Preprocessing)

python
temp_path = output_path.with_suffix('.tmp.nii.gz')
# ... write to temp_path ...
temp_path.rename(output_path)

Grep Cheatsheet

bash
# Find a quality check by ID
grep -n "check_name.*snr\|B1" mengrowth/preprocessing/utils/quality_filtering.py

# Find all blocking checks
grep -n 'action.*=.*"block"' mengrowth/preprocessing/utils/quality_filtering.py

# Find all curation config dataclasses
grep -n "@dataclass" mengrowth/preprocessing/config.py

# Find all preprocessing config dataclasses
grep -n "@dataclass" mengrowth/preprocessing/src/config.py

# Find step execution levels
grep -n "StepMetadata" mengrowth/preprocessing/src/config.py

# Find where a patient gets rejected
grep -rn "mark_excluded\|rejection_reason" mengrowth/preprocessing/utils/

# Find parallel execution points
grep -rn "ProcessPoolExecutor\|ThreadPoolExecutor" mengrowth/

# Find all CLI flags
grep -n "add_argument" mengrowth/cli/curate_dataset.py mengrowth/cli/preprocess.py

# Find brain mask resolution
grep -rn "brain_mask" mengrowth/preprocessing/src/

# Find normalization implementations
grep -rn "class.*Normalizer" mengrowth/preprocessing/src/normalization/

File Format Reference

File Location Schema
rejected_files.csv quality/ source_path, filename, patient_id, study_name, rejection_reason, source_type, stage
quality_issues.csv quality/ patient_id, study_id, modality, file_path, check_name, action, message, level, details
id_mapping.json dataset/ {P*: {new_id: MenGrowth-*, studies: {old_idx: new_id}}}
quality_metrics.json quality/ Hierarchical: patient → study → modality → metric values
per_study_metrics.csv qc_analysis/ One row per (study, sequence) with all computed metrics
metadata_enriched.csv dataset/ patient_id, age, sex, ..., included, exclusion_reason, MenGrowth_ID

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

majiayu000/claude-skill-registry

agent-ops-spec

Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.

163 31
Explore
majiayu000/claude-skill-registry

agent-ops-state

Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.

163 31
Explore
majiayu000/claude-skill-registry

agent-ops-spec

Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.

163 31
Explore
majiayu000/claude-skill-registry

agent-ops-testing

Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.

163 31
Explore
majiayu000/claude-skill-registry

agent-ops-testing

Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.

163 31
Explore
majiayu000/claude-skill-registry

agent-ops-state

Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.

163 31
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results