Skill — text.detect_problem_signal

Intent

Classify whether a piece of text contains a user-expressed problem, pain, or need.

Inputs

detect_request:
  raw_text: string
  platform: string

Outputs

detect_result:
  has_problem_signal: boolean
  signal_markers: string[]
  label: enum [problem, how_to, bug, request, complaint, discussion, showcase, other]
  confidence: number

Tools

• llm.classify

Policy

• Return ONLY valid JSON.
• Never provide advice or solutions.
• signal_markers must be verbatim snippets (no paraphrase).
• If uncertain, return best guess with low confidence.

Failure Modes

• Invalid JSON: caller falls back to heuristic detection.