elicitation_skill

Overview

This skill enables expert agents to perform structured self-review and self-correction on code before final validation. It applies a catalog of 75 elicitation techniques, letting experts choose manually or run an automated selection based on detected code type. The goal is to catch defects early, speed up final validation, and improve the expert's learning loop.

How this skill works

After the Execute phase the skill loads the current context and analyzes the written code to detect its type. Depending on mode, it either prompts the expert with five relevant techniques to pick from, automatically selects and applies techniques, or skips to sniper validation. The skill runs targeted checks, applies fixes, and generates a concise report listing applied techniques and changes.

When to use it

• After writing or modifying code and before sending for final review (sniper).
• When you want to reduce back-and-forth during external validation.
• To enforce domain-specific checks (security, performance, API contracts).
• When you need a repeatable self-review workflow for auditability.
• During refactors to detect regressions and breaking changes.

Best practices

• Always run the Init step to load context and determine mode.
• Use Auto mode for routine code types and Manual mode for complex or novel changes.
• Document every self-correction and list applied techniques in the report.
• Do not apply techniques outside your expertise; escalate when needed.
• Preserve a clear changelog for sniper to speed validation.

Example use cases

• A backend engineer runs Auto mode after adding a new API endpoint to check contracts, error handling, and type coverage.
• A developer refactoring data models uses Manual mode to select migration safety, N+1 detection, and data integrity checks.
• A security-focused change triggers security audit, OWASP checks, and input validation techniques before sniper.
• A UI update is reviewed for accessibility, loading states, and edge-case interactions in Manual mode.
• Performance-critical optimization runs profiling, memory analysis, and complexity checks automatically.