- Industry
- Procure AI
- Department
- Engineering
- Team
- Application Engineering
- Kind
- Decision-maker
- Approver
- Engineering Lead
- Code
- ENG-10
It finds the platform code that costs most to keep, such as per-client approval rule forks, hand-built ERP adapters, legacy match logic and slow spend-report queries, from churn, linked defects, complexity and build time, estimates the saving, and ranks a refactoring backlog by payback. It decides what to refactor next within the capacity set aside for debt and proposes small, behaviour-preserving first changes.
How it works
7 steps, and a person signs off before step 7.
- 01Looks it up
Finds the record the request is about in the code repository: the commit history, complexity metrics, per-client branches and the bugs linked to the files in scope..
- 02Reads the request
Works out what is being asked and what a complete answer needs.
- 03Checks the facts
Checks figures, dates and rules against the source records, not from memory.
- 04Forecasts, when it decides
Projects what each option will cost or return before it recommends one.
- 05Works it through
Prepares the draft, match, recommendation or reply, with its sources.
- 06Asks a person
Hands the work to Engineering Lead with its reasoning before anything goes out.
- 07Acts
Once approved, it acts in the issue tracker: it creates the ranked debt items in the backlog with their estimates and links. and records what was done.
What it works with
Your knowledge, your systems, your people.
Knows
The engineering health standards: complexity and duplication limits, the architecture decision records (configuration over per-client code forks), deprecated libraries and patterns, and the share of capacity set aside for debt.
Reads from
the code repository: the commit history, complexity metrics, per-client branches and the bugs linked to the files in scope.
Acts in
the issue tracker: it creates the ranked debt items in the backlog with their estimates and links.
Hands back
A list ranked by value, each with area (repo and path), cost_signals (an object with churn, defects, complexity, client_forks and build_minutes), proposed_refactor, effort_days, saving_hours_per_sprint, payback_sprints, risk (low, medium or high) and first_change.
Asks a person when
The plan exceeds 20 percent of capacity, changes a public API or module boundary, touches approval, match or payment code, removes a client-specific path still in use, or a change exceeds 300 lines.
Settings
Your limits, set by your managers.
| Setting | Default | What it controls |
|---|---|---|
| Share of capacity for debt (%) | 20 | The most of each sprint's capacity the proposed refactors may take. |
| Hotspot window (days) | 90 | Churn and linked defects are counted over this window. |
| Complexity limit per function | 15 | Functions above this cyclomatic complexity count as a cost signal. |
| Largest refactoring change (lines) | 300 | Each refactoring change stays below this size. |
| Must pay back within (sprints) | 6 | A refactor not repaid in saved time within this many sprints is not proposed. |
Limits are settings, not code: a manager changes them without a developer.
Rules it always follows
- Rank areas by cost: churn and linked defects over 90 days, functions above complexity 15, per-client code branches and build minutes; a cost with no evidence is not counted.
- Prefer replacing a per-client code fork with tenant configuration, as the architecture records direct.
- Propose a refactor only when it pays back within 6 sprints, and keep the plan within 20 percent of capacity.
- Every refactor preserves behaviour, is covered by tests first, and stays below 300 lines; refactors of match, approval or payment code carry control tests that pass before and after.
- Prefer removing patterns the architecture records already retire over introducing new ones.
See the Technical Debt Agent agent on your own work.
We’ll run it on a sample of your work and show you what your approver would see.