AgentSkillsCN

updating-feature-context-from-feedback

根据用户反馈、研究发现或产品变更，更新现有功能内容的上下文。在反馈揭示现有上下文包中的空白、不准确之处，或用户心智模型发生转变时使用。

SKILL.md
--- frontmatter
name: updating-feature-context-from-feedback
description: Updates existing feature content context based on user feedback, research findings, or product changes. Use when feedback reveals gaps, inaccuracies, or shifts in user mental models within an existing context pack.

Updating Feature Context from Feedback

Quick start

Collect or infer:

  • •Existing context pack (or path to it)
  • •Feedback source (user research, support tickets, analytics, PM input)
  • •Feedback type (gap, correction, expansion, deprecation)
  • •Change scope (additive, modifying, removing)

Then produce output using TEMPLATES.md. Validate with RUBRIC.md.

Workflow

  1. •Load the existing context pack and identify its current version.
  2. •Categorize incoming feedback using the feedback taxonomy.
  3. •Map each feedback item to affected context sections (vocabulary, states, actions, constraints).
  4. •Draft proposed changes with explicit before/after deltas.
  5. •Flag conflicts where feedback contradicts existing context or other feedback.
  6. •Generate a change summary with rationale for each modification.
  7. •Update version metadata and changelog.
  8. •Run the rubric check. Revise until it passes.

Degrees of freedom

  • •Default: Low. Context changes must be traceable and justified.
  • •Allowed variation: Rationale phrasing and changelog format may adapt to team conventions, as long as traceability is preserved and rubric passes.

Failure modes to avoid

  • •Applying feedback without validating source credibility
  • •Overwriting context without preserving change history
  • •Conflating user requests with validated needs
  • •Introducing terminology conflicts with existing vocabulary
  • •Removing states without confirming deprecation across all touchpoints

References