Skip to content

Skill Finder: detect repeated prompt/work shapes worth promoting to a skill [R4 from #340] #431

Description

@russmiles

Follow-up R4 from the microsoft/AI-Engineering-Coach exploration (docs/superpowers/explorations/2026-06-16-ai-engineering-coach-findings.md, issue #340).

Idea

The MS Skill Finder detects repeated prompt patterns and proposes promoting them to reusable skills. Our analog — the reflection-driven regression-detection GC rule — is failure-oriented (recurring surprises → proposed constraints). It does not surface reuse-oriented candidates (recurring work shapes → proposed skills).

Scope

  • A detector that groups reflections / task patterns by shape (not just failure theme) and flags 2+ recurrences as skill-authoring candidates.
  • Feeds the skill-authoring pipeline (e.g. component-design-with-tdad) rather than /harness-constrain.
  • Likely a new GC rule or a /reflect-adjacent surface; emits candidates with evidence (dates, the repeated shape) for human disposition.

Constraint

Without session-log data this works over reflections/specs only (artifact substrate). A richer prompt-level version depends on R1's behavioural data.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions