Learns how the user likes to work. By identifying patterns in feedback and corrections, this skill allows the system to preemptively align with the user's preferred style (e.g., "Spec-heavy" vs "Speed-first").

When to use this skill

Continuously in the background of every conversation
When a user repeatedly corrects a specific type of AI behavior
When a user explicitly states a new preference or constraint

Analysis Steps

Track Repeated Constraints: Note if the user always mentions things like "no guess" or "strict types".
Identify Pain Points: Where does the user intervene most? (e.g., "Always edits the design doc").
Detect Dissatisfaction: Analyze negative feedback to identify the underlying skill failure.
Emit Probabilistic Signals: "User prefers AI autonomy: high (confidence 0.8)".

Decision Tree

flowchart TD
    A[Observe Prompt] --> B{Matches Known Pattern?}
    B -->|Yes| C[Reinforce Signal Confidence]
    B -->|No| D{Repeated 3+ Times?}
    D -->|Yes| E[Register New Intent Pattern]
    D -->|No| F[Observe & Wait]
    E --> G[Update Preference Model]
    C --> G
    G --> H[Notify Orchestrator Adapter]

Review Checklist

Evidence: Is there a concrete list of prompts justifying this pattern?
Privacy: Ensure NO personal data or PII is stored in the intent patterns.
Autonomy: Balance intent-following with the project's hard safety rules.
Reversibility: Can the user easily override an inferred preference?

How to provide feedback

Be specific: "The analyzer inferred I prefer 'minimal detail', but I only wanted that for this specific task."
Explain why: "Incorrectly assuming a preference leads to future planning gaps."
Suggest alternatives: "Recommend adding 'Task-Specific' vs 'Permanent' flags to intent signals."

Signals are probabilistic, not absolute.

npx skills add https://smithery.ai