Feature Flag Strategy

Feature Flag Strategy

Overview

Use this skill to manage feature flags as controlled release instruments, not permanent complexity.

Scope Boundaries

• Use this skill when the task matches the trigger condition described in description.
• Do not use this skill when the primary task falls outside this skill's domain.

Shared References

• Lifecycle decision rules:
  • references/flag-lifecycle-decision-rules.md
• Targeting and blast-radius guidance:
  • references/flag-targeting-risk-guidance.md

Templates And Assets

• Flag catalog template:
  • assets/feature-flag-catalog-template.csv
• Rollout plan template:
  • assets/flag-rollout-plan-template.md
• Retirement checklist:
  • assets/flag-retirement-checklist.md

Inputs To Gather

• Rollout objective, user segments, and release risk tolerance.
• Blast-radius constraints and required kill-switch latency.
• Operational/compliance constraints and on-call readiness.
• Flag platform capabilities, dependency behavior, and telemetry coverage.

Deliverables

• Flag definition with owner, type, kill-switch, and expiry.
• Rollout plan with progression gates and rollback policy.
• Lifecycle decision record including retirement criteria.
• Verification evidence for targeting and disable behavior.

Workflow

1. Decide whether a flag is warranted using references/flag-lifecycle-decision-rules.md.
2. Register flags in assets/feature-flag-catalog-template.csv with ownership and expiry.
3. Define rollout and rollback using assets/flag-rollout-plan-template.md.
4. Validate targeting safety using references/flag-targeting-risk-guidance.md.
5. Retire flags using assets/flag-retirement-checklist.md once control is no longer needed.

Quality Standard

• Every flag has owner, kill-switch, and retirement date.
• Rollout and rollback criteria are measurable and pre-defined.
• Targeting rules are auditable and operationally safe.
• Retired flags are removed from code paths, not just disabled.

Failure Conditions

• Stop when flags lack ownership, expiry, or safe disable paths.
• Stop when targeting rules create unbounded or opaque blast radius.
• Escalate when accepted rollout risk exceeds policy thresholds.