parallel-debugging

Parallel Debugging

Framework for debugging complex issues using the Analysis of Competing Hypotheses (ACH) methodology with parallel agent investigation.

When to Use This Skill

• Bug has multiple plausible root causes
• Initial debugging attempts haven't identified the issue
• Issue spans multiple modules or components
• Need systematic root cause analysis with evidence
• Want to avoid confirmation bias in debugging

Hypothesis Generation Framework

Generate hypotheses across 6 failure mode categories:

1. Logic Error

• Incorrect conditional logic (wrong operator, missing case)
• Off-by-one errors in loops or array access
• Missing edge case handling
• Incorrect algorithm implementation

2. Data Issue

• Invalid or unexpected input data
• Type mismatch or coercion error
• Null/undefined/None where value expected
• Encoding or serialization problem
• Data truncation or overflow

3. State Problem

• Race condition between concurrent operations
• Stale cache returning outdated data
• Incorrect initialization or default values
• Unintended mutation of shared state
• State machine transition error

4. Integration Failure

• API contract violation (request/response mismatch)
• Version incompatibility between components
• Configuration mismatch between environments
• Missing or incorrect environment variables
• Network timeout or connection failure

5. Resource Issue

• Memory leak causing gradual degradation
• Connection pool exhaustion
• File descriptor or handle leak
• Disk space or quota exceeded
• CPU saturation from inefficient processing

6. Environment

• Missing runtime dependency
• Wrong library or framework version
• Platform-specific behavior difference
• Permission or access control issue
• Timezone or locale-related behavior

Evidence Collection Standards

What Constitutes Evidence

Evidence Type	Strength	Example
Direct	Strong	Code at file.ts:42 shows if (x > 0) should be if (x >= 0)
Correlational	Medium	Error rate increased after commit abc123
Testimonial	Weak	"It works on my machine"
Absence	Variable	No null check found in the code path

Citation Format

Always cite evidence with file:line references:

**Evidence**: The validation function at `src/validators/user.ts:87`
does not check for empty strings, only null/undefined. This allows
empty email addresses to pass validation.

Confidence Levels

Level	Criteria
High (>80%)	Multiple direct evidence pieces, clear causal chain, no contradicting evidence
Medium (50-80%)	Some direct evidence, plausible causal chain, minor ambiguities
Low (<50%)	Mostly correlational evidence, incomplete causal chain, some contradicting evidence

Result Arbitration Protocol

After all investigators report:

Step 1: Categorize Results

• Confirmed: High confidence, strong evidence, clear causal chain
• Plausible: Medium confidence, some evidence, reasonable causal chain
• Falsified: Evidence contradicts the hypothesis
• Inconclusive: Insufficient evidence to confirm or falsify

Step 2: Compare Confirmed Hypotheses

If multiple hypotheses are confirmed, rank by:

1. Confidence level
2. Number of supporting evidence pieces
3. Strength of causal chain
4. Absence of contradicting evidence

Step 3: Determine Root Cause

• If one hypothesis clearly dominates: declare as root cause
• If multiple hypotheses are equally likely: may be compound issue (multiple contributing causes)
• If no hypotheses confirmed: generate new hypotheses based on evidence gathered

Step 4: Validate Fix

Before declaring the bug fixed:

• Fix addresses the identified root cause
• Fix doesn't introduce new issues
• Original reproduction case no longer fails
• Related edge cases are covered
• Relevant tests are added or updated

Runtime Evidence Pattern

When arbitration leaves hypotheses Plausible or Inconclusive and static evidence is exhausted, inject targeted runtime logs to disambiguate. The pattern is designed so cleanup is mechanical (grep + delete) and the logs never leak into production.

Log Convention

Every injected log MUST follow this format:

[DEBUG] [{file}:{line}] {short description} { {relevant vars} }

Concrete examples per language:

// JS/TS
console.log("[DEBUG] [auth.ts:42] before token verify", { tokenLen: token.length, hasUser: !!user })

# Python
print(f"[DEBUG] [auth.py:42] before token verify", {"token_len": len(token), "has_user": user is not None})
# or
logger.debug("[DEBUG] [auth.py:42] before token verify token_len=%d has_user=%s", len(token), user is not None)

// Rust
eprintln!("[DEBUG] [auth.rs:42] before token verify token_len={} has_user={}", token.len(), user.is_some());

// Go
fmt.Fprintf(os.Stderr, "[DEBUG] [auth.go:42] before token verify token_len=%d has_user=%v\n", len(token), user != nil)

The [DEBUG] prefix is non-negotiable. It is the cleanup anchor.

Strategic Placement

Pick 3-5 points, not more. Flooding logs makes signal harder to extract.

Location	Captures
Function entry	Confirms execution path, args
Before async call	State just before the operation
After async call	Result, error, timing
Conditional	Which branch was taken
Catch block	Error name, message, partial state

What NOT to Log

• Passwords, tokens, API keys (use length or present/absent markers instead)
• PII (email, phone, names) unless masked
• Full request/response bodies (use sizes and selected fields)
• Inside hot loops without rate limiting

Cleanup Protocol

After the bug is verified fixed, remove every [DEBUG] log. The grep is the source of truth — if grep returns zero matches, cleanup is complete.

# JS/TS
grep -rn '\[DEBUG\]' . --include='*.ts' --include='*.tsx' --include='*.js' --include='*.jsx'

# Python
grep -rn '\[DEBUG\]' . --include='*.py'

# Rust
grep -rn '\[DEBUG\]' . --include='*.rs'

# Go
grep -rn '\[DEBUG\]' . --include='*.go'

Multi-line console statements that span more than the matched line must be removed in full. Verify the file still parses after deletion.

Iteration Cap

A debug session should perform at most 2 rounds of log injection. If after the second round the hypotheses are still Inconclusive, escalate to the user with a written summary of what has been ruled out and what additional context (MCP access, production logs, a minimal reproduction) would be needed.

Specialized Investigation Agents

When a hypothesis falls into one of the 6 failure mode categories above, prefer a specialized investigator over a generic team-debugger. The specialized agent loads the right knowledge base automatically and produces higher-precision findings.

Failure mode category	Preferred specialized agent	Notes
Logic Error	senior-review:code-auditor	Failure-flow tracing + pattern consistency
Data Issue	senior-review:code-auditor or senior-review:security-auditor	The latter when the data crosses a trust boundary
State Problem (concurrency, cache, mutation)	senior-review:ui-race-auditor for UI; senior-review:distributed-flow-auditor for cross-service
Integration Failure	senior-review:distributed-flow-auditor	Both sides of the contract
Resource Issue	senior-review:code-auditor + react-development:react-performance-optimizer (if frontend)
Environment	senior-review:chicken-egg-detector	Startup cycles, init order, config bootstrap

team-debugger remains the fallback when no specialized agent matches the hypothesis cleanly, or when the investigation is too cross-cutting for a single specialist.

Sub-spawning caveat

When a team-debugger spawns a specialized sub-agent via the Agent tool (e.g. to deepen one of the 6 categories), that sub-agent itself cannot spawn further sub-agents (Claude Agent SDK restriction: one-level subagent nesting). If a hypothesis requires deeper delegation, the debugger reports this to the team lead rather than chaining indirectly; the lead has the team-level view to decide whether to re-spawn at the top level or escalate to the user.

Output Persistence

The team-debugger agent and the specialized investigators all accept a spawner-provided output file path. When the orchestrator spawns an investigator and wants the structured report on disk (the default in /team-review and similar pipelines), the spawn prompt must include Write your final report to <path>. Investigators write directly with the Write tool rather than returning the report only as message text.

Reference: docs/references/agent-teams-best-practices.md § Operational do's and don'ts.