automated-doubt

name: automated-doubt description: Runs a pipeline of specialized validator subagents post-implementation to audit the code from various critical perspectives (Type Safety, Security, Anxiety/Concurrency, and Public Interfaces).

Automated Doubt Skill

Use this skill after completing code changes, prior to committing or shipping, to perform a rigorous multi-perspective audit.

Workflow

1. Invoke Auditor Subagents Concurrently:

   • Spawn multiple specialized research or self subagents in parallel using the invoke_subagent tool.
   • CRITICAL: The prompts must be goal-oriented, specifying the objective and constraints, leaving the subagents to decide how to analyze the codebase.

   • Type Safety Validator:

     • Objective: Audit the new/modified code for strict type safety.
     • Constraints: Specifically look for loose typings, excessive casting, and forbidden as any expressions (violating user rules).

   • Security Analyst:

     • Objective: Inspect the changes for potential security liabilities.
     • Constraints: Audit for unsanitized inputs, path traversal, injection vectors, and accidental leakage of sensitive environment variables or filesystem paths in logs/errors.

   • Anxiety Reader:

     • Objective: Identify what could go wrong under stress, load, or race conditions.
     • Constraints: Look for concurrent API hits without limit, race conditions, resource/memory leaks, unhandled promise rejections, and missing rate limit handling.

   • Public Interface Validator:

     • Objective: Review modifications to exported APIs, types, or interfaces.
     • Constraints: Verify compatibility, proper encapsulation, and check if change introduces breaking API contracts.

2. Consolidate and Address Findings:

   • Wait for all subagents to finish.
   • Summarize findings into a structured format (categorized by severity: Critical, High, Medium, Low).
   • Address the findings iteratively, re-running the automated doubt validators until quality standards are met.