Qa Tech Debt

Tech Debt Management

Triggers

• "technical debt"
• "tech debt"
• "refactoring priority"
• "legacy code"
• "code quality"

Identification

Code Smells to Detect

# TODOs and FIXMEs
grep -r "TODO\|FIXME\|HACK\|XXX" --include="*.ts" --include="*.tsx" src/

# Large files
find src -name "*.ts" -o -name "*.tsx" | xargs wc -l | sort -n | tail -20

# Complexity (nesting)
grep -r "if.*if.*if" --include="*.ts" src/

# any in TypeScript
grep -r ": any" --include="*.ts" --include="*.tsx" src/

Metrics

| Metric | Threshold | Command | |----------|-------|----------| | LOC/file | 70% | npm test -- --coverage |

Categorization

Impact

| Level | Description | Examples | |--------|-------------|----------| | Critical | Blocks development | Circular coupling | | High | Significantly slows down | Massive duplication | | Medium | Hinders maintenance | Confusing naming | | Low | Cosmetic | Inconsistent style |

Effort

| Level | Time | Examples | |--------|-------|----------| | Trivial | 1 week | Rewrite component |

Prioritization

Impact/Effort Matrix

Impact
  ^
  |  Quick Wins  |  Strategic
  |     (P1)     |    (P2)
  +--------------+-------------
  |   Fill-in    |   Avoid
  |     (P3)     |    (P4)
  +-------------------------> Effort

Remediation Plan

Template

## Item: [Name]

**Priority**: P[1-4]
**Impact**: [Critical/High/Medium/Low]
**Effort**: [Trivial/Low/Medium/High]

### Description
[Description of the problem]

### Files concerned
- path/to/file.ts:L42

### Proposed solution
[Refactoring approach]

### Success criteria
- [ ] Tests pass
- [ ] No regression
- [ ] Improved metrics

Test coverage analysis

Test debt is a first-class category of technical debt. Measure and close the gaps.

# Current coverage (statements / branches / functions / lines)
npm test -- --coverage          # Jest / Vitest

| Metric | Threshold | Notes | |--------|-----------|-------| | Statements | > 80% | | | Branches | > 75% | Hardest and highest-value gap | | Functions | > 80% | |

• Categorise gaps by criticality (business code and edge cases first), not by raw %.
• Prioritise critical business code; 100% coverage != 100% quality — never sacrifice test

quality to hit a number.

• Add missing tests for branches, boundary conditions and error paths; wire a coverage

gate into CI (/ops:ops-ci). Generate the tests via /dev:dev-tdd or /dev:dev-tdd.

Continuous improvement (Kaizen)

For incremental, durable improvement of code and process, run the PDCA cycle instead of a big-bang rewrite.

• PLAN: identify the problem and root cause (5 Whys), set a SMART objective.
• DO: implement one change at a time, atomic commits.
• CHECK: measure before/after, compare to the objective.
• ACT: standardise on success, adjust on failure; plan the next iteration.
• Eliminate the 7 Muda (overproduction, waiting, transport, over-processing, inventory,

motion, defects). Always measure before and after; one change at a time, no revolutions.

Workflow

1. Identify - Scan the codebase (smells, coverage gaps, Muda)
2. Categorize - Impact and effort
3. Prioritize - Decision matrix
4. Plan - Integrate into the backlog
5. Execute - Incremental refactoring (one change at a time, PDCA)
6. Validate - Tests, coverage and metrics (before/after)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: christopherlouet
• Source: christopherlouet/claude-base
• License: MIT
• Homepage: https://christopherlouet.github.io/claude-base/

Install and usage instructions live in the source repository linked above.