Feature complexity audit

Estimate the permanent cost of a feature — cognitive, technical, operational, and governance carry — before the build commitment, producing a complexity scorecard and a reduction recommendation.

Canonical skill profile. This page is generated from the repository contract. The linked SKILL.md remains the source of authority.

Jump to

• Overview
• When to use
• When not to use
• Core moves
• Expected output
• Verification
• Handoff signals
• Pairs well with
• Anti-patterns

View the canonical contract on GitHub.

Overview

Use this skill to make the permanent cost of a feature visible before committing to build. A feature is a bet that consumes complexity forever; this skill estimates that carry across cognitive, technical, operational, and governance dimensions, and recommends the cheapest coherent way to get the value. It produces a coarse, honest scorecard — never a false-precision number.

When to Use

• A feature looks valuable and complexity is becoming the deciding factor.
• Before a build_now verdict on anything non-trivial.
• When comparing two designs that deliver similar value at different carry.

When NOT to Use

• The change is tiny, local, and obviously cheap.
• Value itself is unresolved (resolve the lever first).
• No scope or dependencies are known yet (nothing to estimate).

Core Moves

1. Frame. State the feature and the scope being costed.
2. Score four dimensions (each low/medium/high with drivers):
   • Cognitive — surface users and the team must hold in their heads.
   • Technical — new dependencies, data model, integration, migration surface.
   • Operational — support, on-call, runbook, monitoring carry.
   • Governance — security, privacy, compliance, accessibility obligations.
3. Name reversibility. reversible / partially_reversible / one_way_door, and the mechanisms (flag, cohort, rollback, sunset, migration plan).
4. Roll up to a coarse level (low/medium/high) with the dominant driver named.
5. Recommend reduction. The smallest scope that keeps the value, or the explicit exception required if the high cost is accepted.

Optional Modules

• Carry-over forecast — Estimate the ongoing yearly cost (support tickets, on-call, doc upkeep), not just the build cost.
• Reduction options — List 1–3 concrete ways to cut permanent cost without killing the value.
• One-way-door check — When reversibility is low, force a technical gate before commit.

Activation Triggers

• Run whenever complexity is the pivotal factor in a worth-doing verdict.
• Use the one-way-door check when reversibility is one_way_door.
• Use reduction options whenever the rolled-up level is high.

Inputs (minimum)

• Feature, scope, dependencies, UX impacted, operational footprint.

Outputs (minimum)

• A permanent-cost score, risks, and a reduction recommendation — mapping to the complexity_cost and reversibility fields of the Feature Value Governance Contract. A high level under a build_now verdict requires decision.exception_approval.

Expected Output

complexity_scorecard:
  feature: <one line>
  dimensions:
    cognitive: { level: low|medium|high, drivers: <text> }
    technical: { level: low|medium|high, drivers: <text> }
    operational: { level: low|medium|high, drivers: <text> }
    governance: { level: low|medium|high, drivers: <text> }
  reversibility:
    level: reversible | partially_reversible | one_way_door
    mechanisms: [ ... ]
  rolled_up_level: low | medium | high
  dominant_driver: <text>
  reduction_recommendation: <smallest scope that keeps value, or required exception>
  uncertainty: <what is hard to estimate>

Verification

• All four dimensions are scored, not just the technical one.
• Reversibility is explicit.
• The recommendation reduces cost or names the exception — it does not hand-wave.
• The level is coarse and honest, not a fake decimal.

Handoff Signals

• A high level under a build verdict → require exception_approval and route to human review.
• The feature should perhaps not be built at all → feature-value-governance.
• Structural design is the real question → architecture-review.

Pairs Well With

• feature-value-governance
• architecture-review
• premortem

Anti-patterns

• Scoring only build effort and ignoring forever-carry.
• A precise-looking number that hides deep uncertainty.
• Treating high complexity as an automatic veto, or as no cost at all.