Reclassify Tests
SkillDev toolsReclassify tests by adding @pytest.mark.functionality to tests not explicitly shown in the spec. Invoke with /reclassify-tests <problem> <checkpoint>.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Reclassify Tests skill
What this skill tells your AI
The instructions your AI receives, as published by sprocketlab/slop-code-bench in .claude/skills/reclassify-tests/SKILL.md and read by ahel’s review.
Analyze a checkpoint's spec and test file to identify tests that are NOT explicitly demonstrated in the spec examples, and add @pytest.mark.functionality markers to them.
Usage: /reclassify-tests file_backup checkpoint_1
Purpose
Core tests (unmarked) should only cover behaviors that are explicitly demonstrated in spec examples. Tests that verify implicit requirements, edge cases, or inferred behaviors should be marked as @pytest.mark.functionality.
This skill helps ensure fair evaluation by:
- Keeping core tests limited to explicit spec examples
- Moving implicit/inferred behaviors to functionality tests
- Making test classification consistent and justifiable
Step 1: Gather Context
Read these files for the specified problem/checkpoint:
problems/{problem}/checkpoint_N.md # Spec with examples
problems/{problem}/tests/test_checkpoint_N.py # Test file to reclassify
problems/{problem}/tests/conftest.py # Fixtures context
problems/{problem}/tests/data/checkpoint_N/ # Test case data (if present)
Step 2: Extract Explicit Spec Behaviors
From the spec, identify behaviors that are explicitly demonstrated:
What Counts as "Explicit"
- Direct Examples: Code blocks showing exact input → output
- Explicit Statements: "MUST", "SHALL", specific values defined
- Example Output: JSON/text showing exact expected format
- Stated Requirements: Numbered/bulleted requirements with specific behavior
What Does NOT Count as "Explicit"
- Implied Behavior: Reasonable inferences not shown in examples
- Edge Cases: Boundary conditions not demonstrated
- Error Handling: Unless specific errors are shown in examples
- Format Details: Exact formatting beyond what examples show
- Ordering: Unless explicitly stated or shown in all examples
Create a Checklist
For each spec example, note:
- What input is shown?
- What output is shown?
- What behavior is demonstrated?
Step 3: Analyze Each Test
For each test function or parametrized test case:
A. Identify What the Test Verifies
- What specific behavior does this test check?
- What input does it use?
- What output does it expect?
B. Find Spec Support
Ask: "Can I point to a spec example that explicitly demonstrates this exact behavior?"
| Spec Support | Classification |
|---|---|
| Example shows exact scenario | Keep as core (no marker) |
| Behavior stated but no example | Mark as functionality |
| Inferred from context | Mark as functionality |
| Edge case not shown | Mark as functionality |
| Error handling not in examples | Mark as error |
C. Document the Decision
For each test, note:
- Spec support (quote or "NONE")
- Classification decision
- Reason
Step 4: Apply Markers
For Tests to Reclassify
Add @pytest.mark.functionality ABOVE the test function or parametrize decorator:
Before:
@pytest.mark.parametrize("case", EDGE_CASES, ids=[c["id"] for c in EDGE_CASES])
def test_edge_cases(entrypoint_argv, case, tmp_path):
...
After:
@pytest.mark.functionality
@pytest.mark.parametrize("case", EDGE_CASES, ids=[c["id"] for c in EDGE_CASES])
def test_edge_cases(entrypoint_argv, case, tmp_path):
...
For Individual Parametrized Cases
If some cases in a parametrized test are explicit and others aren't, consider:
- Split the test: Separate core cases from functionality cases
- Use pytest.param with marks:
@pytest.mark.parametrize("case", [
pytest.param(case1, id="explicit-case"),
pytest.param(case2, marks=pytest.mark.functionality, id="inferred-case"),
])
def test_cases(case):
...
Step 5: Report Changes
Generate a summary of changes made:
# Reclassification Report: {problem} {checkpoint}
## Summary
- Tests analyzed: N
- Tests reclassified: N
- Tests unchanged: N
## Reclassified Tests
### `test_function_name` → @pytest.mark.functionality
**Reason**: Test verifies [behavior] which is not explicitly demonstrated in spec.
**Spec says**: [Quote or "No example shown"]
**Test expects**: [What the test checks]
---
### `test_another_function` → @pytest.mark.functionality
...
## Unchanged Tests (Explicitly in Spec)
| Test | Spec Example Reference |
|------|------------------------|
| `test_core_case_1` | Example 1: shows exact input/output |
| `test_core_case_2` | Example 2: demonstrates this scenario |
Classification Guidelines
KEEP AS CORE (no marker) when:
- Spec example shows exact input and expected output
- Behavior is explicitly stated with specific values
- Test directly validates a demonstrated scenario
- Example output format matches test expectations exactly
MARK AS FUNCTIONALITY when:
- Behavior is reasonable but not shown in examples
- Test checks edge cases not demonstrated
- Test verifies format details beyond examples
- Test validates inferred requirements
- Test checks ordering not explicitly stated
- Test validates error messages/codes not shown
MARK AS ERROR when:
- Test validates error handling for invalid input
- Test checks failure modes
- Test expects non-zero exit or exceptions
Common Reclassification Scenarios
1. Implicit Ordering
Spec: "Process all jobs"
Test: Expects specific alphabetical order
→ Mark as functionality (unless spec explicitly shows ordering)
2. Edge Values
Spec: "Duration in hours"
Test: Checks duration=0, duration=MAX
→ Mark as functionality (unless 0/MAX shown in examples)
3. Format Details
Spec: Shows JSON output example
Test: Checks exact whitespace, key ordering
→ Mark as functionality (unless spec says "must match exactly")
4. Error Handling
Spec: "Return non-zero on error"
Test: Checks specific error code or message
→ Mark as error (and consider relaxing the assertion)
5. Field Presence
Spec: Example shows fields A, B, C
Test: Requires field D not in any example
→ Mark as functionality
Data-Driven Test Considerations
Many test files use data directories like:
tests/data/checkpoint_N/
├── core/ # Core test cases
├── hidden/ # Functionality test cases
├── errors/ # Error test cases
When test data is already organized this way:
-
Check that markers match data directories:
- Tests loading from
core/should have no marker - Tests loading from
hidden/should have@pytest.mark.functionality - Tests loading from
errors/should have@pytest.mark.error
- Tests loading from
-
Consider moving cases between directories if:
- A "core" case tests implicit behavior → move to
hidden/ - A "hidden" case tests explicit behavior → move to
core/
- A "core" case tests implicit behavior → move to
Example Analysis
Given spec excerpt:
## Example 1
Input: `--now 2025-09-10T03:30:00Z --duration 1`
Output:
{"event":"SCHEDULE_PARSED","timezone":"UTC","jobs_total":2}
{"event":"JOB_ELIGIBLE","job_id":"job-root","kind":"daily",...}
And tests:
def test_schedule_parsed_event():
"""Verify SCHEDULE_PARSED is emitted first."""
# ✓ KEEP AS CORE - Example 1 shows this event first
def test_schedule_parsed_fields():
"""Verify SCHEDULE_PARSED has timezone and jobs_total."""
# ✓ KEEP AS CORE - Example 1 shows these exact fields
def test_empty_schedule():
"""Verify empty schedule returns jobs_total=0."""
# → MARK AS FUNCTIONALITY - No example shows empty schedule
def test_invalid_timezone():
"""Verify invalid timezone returns error."""
# → MARK AS ERROR - Error handling case
After Reclassification
-
Run tests to ensure markers are syntactically correct:
cd problems/{problem} && pytest --collect-only tests/test_checkpoint_N.py -
Verify counts match expectations:
pytest --collect-only -m "not functionality and not error" tests/ # Core only pytest --collect-only -m functionality tests/ # Functionality only pytest --collect-only -m error tests/ # Error only -
Review the changes - did any truly explicit tests get marked?
Signals
- GitHub stars
- 185
- Forks
- 34
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
reclassify-tests- Source
- github.com/sprocketlab/slop-code-bench