---
id: fault-localization-certificate
kind: task-template
title: Fault Localization Certificate
summary: Semi-formal certificate for localizing the most likely buggy region from failing-test behavior.
category: reasoning
status: stable
version: 1
source_alignment: appendix-b
supported_targets:
  - codex
  - claude-code
supported_languages:
  - any
owner: core
generated: false
---

# Fault Localization Certificate

This certificate preserves the paper's four-phase structure:

1. test semantics analysis
2. code path tracing
3. divergence analysis
4. ranked predictions

## Use when

- a failing test exists but the buggy line is unknown
- the failure manifests in one place but may originate elsewhere
- you need ranked candidate regions with evidence

## Inputs

- failing test name
- failing test code
- available source files
- any known passing tests
- optional stack signal or symptom description

## Required rule

Every ranked prediction must trace back through a **CLAIM** and that claim must trace back to a **PREMISE**.

## Certificate

### Phase 1 - test semantics analysis

- What does the failing test do step by step?
- What are the explicit assertions or expected exceptions?
- What is the expected behavior versus the observed failure mode?

#### Formal premises

- PREMISE T1:
- PREMISE T2:
- PREMISE T3:
- PREMISE T4:

### Phase 2 - code path tracing

Build the path from test entry point into production code.

#### Call sequence

| Step | Method / Function | File:Line | Behavior | Why Relevant |
|---|---|---|---|---|
| 1 | `[symbol]` | `[file:line]` | `[what it does]` | `[relation to test]` |
| 2 | `[symbol]` | `[file:line]` | `[what it does]` | `[relation to test]` |

#### Significant observations

- O1:
  - location:
  - observation:
  - why it matters:
- O2:
  - location:
  - observation:
  - why it matters:

#### Indirection check

- registration / factory / config / metadata path involved?:
- hidden collaborator or adapter involved?:
- state mutation path involved?:
- multiple files likely involved?:

### Phase 3 - divergence analysis

For each candidate divergence between expected and actual behavior:

- CLAIM D1:
  - location:
  - code behavior:
  - which premise it contradicts:
  - why:
  - evidence:
- CLAIM D2:
  - location:
  - code behavior:
  - which premise it contradicts:
  - why:
  - evidence:

### Competing hypotheses

- H1:
  - statement:
  - supporting evidence:
  - contradictory evidence:
  - status: plausible | weakened | refuted
- H2:
  - statement:
  - supporting evidence:
  - contradictory evidence:
  - status:
- H3:
  - statement:
  - supporting evidence:
  - contradictory evidence:
  - status:

### Phase 4 - ranked predictions

Each prediction must cite the supporting claim(s).

#### Rank 1

- location:
- candidate region:
- why it is likely:
- supported by claims:
- confidence:
- root cause or crash site?:

#### Rank 2

- location:
- candidate region:
- why it is likely:
- supported by claims:
- confidence:
- root cause or crash site?:

#### Rank 3

- location:
- candidate region:
- why it is likely:
- supported by claims:
- confidence:
- root cause or crash site?:

#### Rank 4

- location:
- candidate region:
- why it is likely:
- supported by claims:
- confidence:
- root cause or crash site?:

#### Rank 5

- location:
- candidate region:
- why it is likely:
- supported by claims:
- confidence:
- root cause or crash site?:

### Exploration log linkage

Link the live exploration record here:

- exploration log path:
- most important confirmed hypothesis:
- most important refuted hypothesis:
- unresolved uncertainty still affecting ranking:

### Alternative root-cause check

If Rank 1 is wrong, which other candidate best explains the failure, and what evidence would decide between them?

- alternative candidate:
- distinguishing evidence needed:
- current status:

### Formal conclusion

- Most likely root cause:
- Best alternative:
- Why the top prediction outranks the crash site or symptom location:
- Final ranked list:
- Evidence completeness:
- unresolved blind spots:

## Structured exploration format

Use `templates/logs/semiformal-exploration-log.md` while filling this certificate.
Do not retroactively invent the exploration record.

## Purpose

Provide a structured, evidence-backed certificate for localizing the most likely buggy region from failing-test behavior.

## When to Use

Use when a failing test exists but the buggy line is unknown, when the failure manifests in one place but may originate elsewhere, or when ranked candidate regions with evidence are needed.

## Inputs

- Failing test name and code
- Available source files
- Any known passing tests

## Optional Inputs

- Stack signal or symptom description
- Related exploration logs

## Constraints

- Every ranked prediction must trace back through a CLAIM and that claim must trace back to a PREMISE
- The exploration log must be written live, not retroactively

## Expected Outputs

- Completed fault localization certificate with ranked predictions
- Formal conclusion with most likely root cause and ranked list

## Acceptance Criteria

- Test semantics are analyzed
- Code paths are traced from test entry points
- Divergence analysis links claims to premises
- Ranked predictions cite supporting claims

## Quality Gates

- Evidence completeness is stated
- Alternative root-cause check is present
- Exploration log is linked

## Suggested Workflow

1. Analyze test semantics (Phase 1)
2. Trace code paths (Phase 2)
3. Perform divergence analysis (Phase 3)
4. Rank predictions (Phase 4)
5. Link exploration log
6. Check alternative root cause
7. Write formal conclusion

## Related Commands

_No specific CLI commands required._

## Related Agents

- TDD guide agent
- Build error resolver agent

## Examples

_Refer to the certificate template sections above for field-level examples._