spec-kit-feedback

spec-kit-feedback is a support skill for spec-kit workflows that captures usage patterns and session observations so prompts and templates can be improved over time. Its main purpose is to help teams understand where workflows succeed, where they create confusion, and where expected quality gates may not fit real usage. Rather than changing the workflow itself, it focuses on noticing completion events, friction during execution, and quality-related deviations, then turning those observations into feedback that can be shared with the spec-kit team or used for RL intake submissions.

The skill activates in several defined situations: when a workflow is completed, when friction appears during the conversation, when users mention problems with prompts or templates, and when quality gates are skipped or become problematic. During a session, it keeps mental notes on workflow type, clarification count, retry count, gate compliance, and completion status. It also watches for patterns such as user confusion, repeated retries, skipped template sections, added content outside the template, or cases where the user takes over a step. At the end of a session, it can provide a structured summary with status, estimated duration, observations, friction points, and positive patterns. It also reminds users to export chat logs and includes export guidance for Claude Code, along with notes on preparing materials for RL intake.

This skill is useful for teams maintaining or improving spec-kit-based workflows, especially those who want evidence about prompt quality, template fit, and workflow usability. It is aimed at users completing spec-kit reviews or related workflows, as well as maintainers who analyze workflow behavior across sessions. It is particularly relevant when workflows need optimization based on real usage rather than assumptions.