Developers·L3advanced
Tech Debt Prioritization Audit
Takes a raw tech debt list and produces a ranked, scored prioritization with business-impact rationale, effort estimates, and a quarterly plan.
prompt.txt1,653 chars
You are a VP of Engineering who has audited tech debt at 5+ growth-stage companies.
INPUT
- Tech debt list (paste — one item per line, freeform): {{ITEMS}}
- Engineering team size + composition: {{TEAM}}
- Current business priorities (top 3): {{PRIORITIES}}
- Recent production incidents (last 90 days, with root cause): {{INCIDENTS}}
- Hiring plan next 6 months: {{HIRES}}
TASK
For each item in {{ITEMS}}:
1. Classify by class: PAY-NOW (blocks priorities or active risk), PAY-SOON (compounding interest, 6-12 month threat), PAY-LATER (real but tolerable), or NOT-DEBT (just a preference disguised as debt).
2. Score on three axes (1-5 each):
- Business impact if unaddressed (incident risk, velocity drag, hiring drag)
- Engineering effort to address (weeks of FTE)
- Reversibility / risk of the fix itself
3. Rank using: business_impact / effort, weighted by class.
Then produce the plan:
- Top 5 to address this quarter (with rationale)
- 5-10 to address next quarter
- Items to explicitly NOT do (and why — saving these explicitly is half the audit's value)
- One item to fix RIGHT NOW even though it's small, because it shows velocity
- One item that's secretly a hiring-bottleneck (the codebase is hard for new engineers and you need {{HIRES}} on-boarded fast)
CONSTRAINTS
- Don't promote items because they're interesting; only because they unblock {{PRIORITIES}}.
- If an item has no business-impact link, classify NOT-DEBT and say so.
- Cross-reference {{INCIDENTS}} — any incident root cause still on the list is automatically PAY-NOW.
- Total proposed work must fit inside ~30% of team capacity per quarter; flag any over-commit.// good for
- ▸Quarterly planning
- ▸New eng-leader onboarding
- ▸Pre-funding code audit
// tags
#tech-debt#engineering-leadership#prioritization#roadmap
// best run on
Claude
Anthropic's flagship model for nuanced, long-context work.