Skip to content

borghei/Claude-Skills / engineering/tech-debt-tracker

tech-debt-tracker

Verified

Scan codebases for technical debt with AST parsing, prioritize by impact, and generate trend dashboards. Use when tracking tech debt, prioritizing refactoring, calculating cost-of- delay, planning sprint debt, or reporting debt to execs.

Tech Debt Tracker

The agent identifies, scores, prioritizes, and tracks technical debt across codebases using AST parsing, cost-of-delay analysis, and trend dashboards.

Core Capabilities

  • Detection — AST parsing (Python) and regex pattern matching (all languages) across six debt categories: code, architecture, test, documentation, dependency, infrastructure.
  • Interest rate — the number this skill adds: interest = impact x frequency (how much the debt slows work each time the code is touched), and cost of delay = interest x sprints x team multiplier. Rank by cost of delay over effort; WSJF and RICE are available as alternatives.
  • Data flow — debt_scanner.py produces the inventory, debt_prioritizer.py scores and schedules it, debt_dashboard.py compares snapshots over time.

When to Use

  • Tracking and quantifying technical debt across a repository.
  • Prioritizing refactoring work and calculating cost-of-delay.
  • Planning sprint capacity allocation between debt and features.
  • Reporting debt health, trends, and investment recommendations to execs.

Clarify First

Before scanning or reporting, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.

  • Target codebase: the directory to scan (the subject of the debt inventory).
  • Prioritization framework & team size: cost-of-delay / WSJF / RICE and headcount (--framework, --team-size; changes the ranking and sprint allocation).
  • Report audience: exec dashboard vs engineering inventory (sets the report format and altitude).

Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.

Tools

ToolPurposeCommand
debt_scanner.pyScan a directory for debt signals; output JSON inventory + text reportpython scripts/debt_scanner.py <dir> --output scan_results --format both
debt_prioritizer.pyEnrich inventory with cost-of-delay/WSJF/RICE and sprint allocationpython scripts/debt_prioritizer.py scan_results.json --framework wsjf --team-size 8
debt_dashboard.pyTrend analysis, velocity, forecasts, and exec summary across snapshotspython scripts/debt_dashboard.py --input-dir ./debt_scans/ --period quarterly

Runnable on the bundled samples:

python scripts/debt_scanner.py assets/sample_codebase --format text
python scripts/debt_prioritizer.py assets/sample_debt_inventory.json --framework cost_of_delay
python scripts/debt_dashboard.py assets/historical_debt_2024-01-15.json assets/historical_debt_2024-02-01.json

The scanner reports each duplicated region once (overlapping windows are merged and repetitive runs inside one file are ignored), so the 986-line sample yields about 60 items. Health score is 100 - severity-weighted items per 1,000 lines (critical 5, high 3, medium 1, low 0.5), so it compares codebases of different sizes. Exit codes: 0 ok; 2 bad input (missing directory or file, unreadable inventory).

References

Load the reference that matches the task — keep this file lean and pull detail on demand:

  • references/methodology.md — the 7-step workflow, debt-classification table, severity scoring framework, interest-rate/cost-of-delay formulas, prioritization matrix, WSJF, sprint allocation ratios, the debt-item JSON schema, refactoring strategies, and quarterly planning. Read when scoring, prioritizing, or planning.

  • references/tool-reference.md — full parameter tables, examples, and output-format details for all three scripts plus the troubleshooting table. Read when running the scripts or debugging output.

  • references/dashboards-and-examples.md — executive and engineering dashboard layouts, a worked Python-microservice scan example, and the success-criteria bar. Read when generating reports or validating quality.

  • references/debt-classification-taxonomy.md — comprehensive taxonomy for classifying debt across dimensions with detection heuristics per category. Read when calibrating detection or labeling items.

  • references/prioritization-framework.md — deep prioritization approaches based on business value, risk, effort, and strategic alignment. Read when designing a prioritization rubric.

  • references/stakeholder-communication-templates.md — templates and guidelines for communicating debt status, impact, and recommendations to different stakeholder groups. Read when reporting to execs or product.

  • references/quadrant-and-roadmap.md — the technical debt quadrant and the implementation roadmap phases. Read when framing debt for a team new to tracking it.

Scope & Limitations

This skill covers:

  • Static detection of code-level, architecture, test, documentation, dependency, and infrastructure debt via AST parsing (Python) and regex pattern matching (all languages).
  • Quantitative prioritization of debt items using cost-of-delay, WSJF, and RICE frameworks with configurable team size and sprint capacity.
  • Historical trend analysis, health scoring, debt velocity tracking, and executive/engineering dashboard generation from multiple scan snapshots.
  • Sprint allocation planning with capacity-aware backlog scheduling and effort estimation by debt type.

This skill does NOT cover:

  • Runtime performance profiling or production monitoring -- see engineering/performance-profiler and engineering/observability-designer for those concerns.
  • Dependency vulnerability scanning (CVE detection) or software composition analysis -- see engineering/dependency-auditor for security-focused dependency review.
  • Automated refactoring or code transformation -- the skill identifies and prioritizes debt but does not modify source code.
  • Database schema debt, API contract drift, or infrastructure-as-code drift detection -- see engineering/database-schema-designer, engineering/api-design-reviewer, and engineering/migration-architect for those domains.

Integration Points

SkillUse for
engineering/dependency-auditorCVE and outdated-dependency findings to merge into the inventory
engineering/ci-cd-pipeline-builderGate builds on summary.health_score or new critical items

Anti-Patterns

Counting Items Instead of Interest

Mistake: Reporting "we have 412 debt items" and prioritising by count per module. Why it happens: Counts are what scanners produce. Instead: Rank by interest: debt in code touched every sprint costs far more than debt in a file nobody opens. Weight by change frequency (for example git log --since=6.months --name-only).

A Debt Sprint Every Quarter

Mistake: Saving all debt work for a dedicated cleanup sprint. Why it happens: It is easy to negotiate once. Instead: Reserve a fixed share of every sprint (the prioritizer's allocation) and fix debt where feature work already touches the code.

Dashboards Without a Decision

Mistake: Sending executives a health trend with no ask. Why it happens: The dashboard looks like the deliverable. Instead: Lead with the decision: what to fund, what it unblocks, and what happens to delivery speed if nothing changes.

Files

FilePurpose
scripts/debt_scanner.pyCodebase scan to a debt inventory with health score
scripts/debt_prioritizer.pyCost-of-delay, WSJF or RICE ranking and sprint allocation
scripts/debt_dashboard.pyTrends, velocity and forecasts across snapshots
assets/sample_codebase/Three-file codebase with planted debt
assets/sample_debt_inventory.json, historical_debt_*.jsonInventory and snapshots for the prioritizer and dashboard
expected_outputs/*.jsonReference outputs for the samples
references/*.mdMethodology, tool reference, dashboards, taxonomy, prioritization, communication templates, quadrant and roadmap
planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies.Stars: 108kUpdated Claude CodeCodexVerifiedInstall
meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要".Stars: 108kUpdated Claude CodeCodexVerifiedInstall