Sunset decision

Decide whether an existing feature should be kept, limited, refactored, deprecated, or removed — based on usage, cost, support load, strategic alignment, and dependencies — and produce an auditable sunset decision record.

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 judge an existing feature against its ongoing cost. Most governance looks forward at new bets; this one looks back at what is already carried. It weighs usage, permanent cost, support load, strategic alignment, and dependencies, and returns one of five dispositions with a migration/removal plan when warranted. Killing a feature that no longer earns its keep is a healthy outcome, not a failure.

When to Use

• A feature has low traction but ongoing maintenance, support, or security cost.
• A periodic portfolio review of carried features.
• A feature blocks a platform upgrade or widens the attack surface.
• Cost-cutting pressure needs a principled, auditable way to choose what to drop.

When NOT to Use

• The feature is new and has not had a fair chance to show traction.
• The decision is about a new feature (use feature-value-governance).
• There is no usage, cost, or support data at all to reason from.

Core Moves

1. Frame and resolve knowledge. Name the feature and pull the signals: usage, cost, support load, strategic alignment, dependencies. Resolve strategic_framework if present.
2. Read traction vs. cost. Is value-per-cost still positive, flat, or negative?
3. Map dependencies. Who/what breaks if it changes — internal and customer-facing.
4. Choose a disposition: keep, limit, refactor, deprecate, or remove.
5. Attach a plan. For deprecate/remove: a migration/removal plan with a churn guardrail. For limit/refactor: the reduced scope. For keep: the next review trigger.
6. State evidence and uncertainty. Cite the data; name what is not known (e.g. strategic accounts whose migration friction is unsized).

Optional Modules

• Migration plan — Phased deprecation: announce → provide path → grace period → decommission, with a churn guardrail.
• Strategic-account check — Identify high-value accounts that depend on the feature and size their migration friction before removal.
• Cost-recovery estimate — Quantify the maintenance/security/upgrade cost recovered by removal.

Activation Triggers

• Always pull usage AND cost AND support signals; a disposition with only one is unsafe.
• Use the strategic-account check before any remove or deprecate disposition.
• Use the migration plan whenever the disposition is deprecate or remove.

Inputs (minimum)

• Usage, cost, support load, strategic alignment, dependencies.

Outputs (minimum)

• One disposition (keep / limit / refactor / deprecate / remove) with a plan and an auditable rationale — recorded as a sunset (or park) decision on the Feature Value Governance Contract.

Expected Output

sunset_decision_record:
  feature: <one line>
  mode: generic | knowledge_aware
  signals:
    usage: <summary + numbers>
    permanent_cost: <summary>
    support_load: <summary>
    strategic_alignment: <summary>
    dependencies: [ ... ]
  value_per_cost: positive | flat | negative
  disposition: keep | limit | refactor | deprecate | remove
  plan: <migration/removal plan, reduced scope, or next review trigger>
  churn_guardrail: <what must not happen>
  evidence: [ ... ]
  uncertainty: <what is not yet known>

Verification

• The disposition rests on usage AND cost AND support, not a single signal.
• Deprecate/remove always carries a migration plan and a churn guardrail.
• Strategic-account dependencies are checked before removal.
• Evidence and uncertainty are both stated.

Handoff Signals

• A removal needs a structural plan → architecture-review.
• The feature should be reshaped rather than removed → feature-planning.
• The portfolio question is really “what new bet replaces this” → feature-value-governance.

Pairs Well With

• feature-value-governance
• feature-complexity-audit
• observability-review

Anti-patterns

• Keeping a feature because removing it feels like admitting failure.
• Removing a feature without a migration path for the accounts that use it.
• Deciding on usage alone while ignoring the cost it offsets (or vice versa).