Agents / Agents

Tech Debt Surveyor

Agent

Turns 'the codebase is a mess' into a managed register: every significant debt named, priced by what it costs you monthly, and ranked — so paydown becomes a plan instead of a complaint.

Best for

  • Teams that discuss debt every retro and pay none of it
  • Making the business case for refactoring in money and velocity
  • Deciding which mess matters and which can rot in peace

What you give it

  • The codebase, its history, and where delivery currently hurts
  • The roadmap, so debt is ranked against where work will actually happen

What you get back

  • A debt register: each item with location, kind, and the harm it does today
  • Interest estimates: what each debt costs monthly in slowdown, bugs and risk
  • A ranked paydown plan woven into the roadmap — plus the explicit 'leave it' list

How it works

  1. Surveys with evidence: change history (what is slow and bug-prone to touch), complexity hotspots, test blind spots, dependency age, and the places developers detour around.
  2. Names each debt precisely: what it is, where, what kind — design, test, dependency, knowledge, duplication.
  3. Estimates interest honestly: the recurring monthly cost in slower changes, escaped bugs and risk exposure.
  4. Ranks by interest x roadmap collision: expensive debt in code the roadmap will visit outranks cheap debt in a corner nobody enters.
  5. Writes the plan as roadmap items with before/after measures — and defends the 'leave it' list, because paydown has opportunity cost too.

Example

You: Every estimate doubles when a task touches the billing module. Survey our debt before we plan next quarter.

Result: A register of 23 items. Top three by interest: the billing module's tangled state (touched by 60% of roadmap items — refactor scheduled before them), a test suite blind spot that let four regressions through this quarter, and a two-versions-behind framework blocking a security patch. Also: nine items explicitly marked 'leave it' — stable, isolated, not worth the risk of touching.

Limits — please read

  • Interest estimates are structured judgement, not accounting; it shows the evidence behind each.
  • The register reflects the code's reality; organisational debt (knowledge silos, process) is noted where visible but owned by you.
  • A register unmaintained becomes debt itself — it sets the review rhythm so the register stays alive.