AgentSkillsCN

eng-iterative-delivery

通过功能标志、回滚机制,以及内置的沟通渠道,将工作分解为安全可控、可观察的增量步骤。

SKILL.md
--- frontmatter
name: eng-iterative-delivery
description: Decompose work into safe, observable increments with feature flags, rollbacks, and communication baked in.

Iterative Delivery and Risk Management

Intent

  • Ship value continuously while containing blast radius.
  • Keep work reviewable, reversible, and traceable across repos and chains.

Workflow

  1. Slice intentionally
    • Break work into minimal vertical slices delivering user-facing or foundational value.
    • Capture dependencies and sequencing so parallel work stays unblocked.
  2. Plan release guards
    • Decide on feature flags, staged rollouts, or canary networks.
    • Define rollback playbooks (contract pause switch, mobile kill switch, server config toggle).
  3. Track progress transparently
    • Maintain an up-to-date todo or checklist tied to commits/PRs.
    • Communicate blockers early; keep reviewers informed of scope changes.
  4. Automate deployment confidence
    • Ensure CI covers each slice; prefer trunk-based workflows with short-lived branches.
    • For migrations, include forward + backward compatibility steps in the same change when feasible.

Verification

  • Each slice has independent validation and observability; no “big bang” merges.
  • Rollback steps documented alongside the change (flags, scripts, transactions).
  • Release notes or deployment checklist updated before handoff to operations/reviewers.