/ak:autoresearch "Compare migration strategies for auth session storage before implementation; note assumptions, affected systems, and the decision needed"

Use when:

Use when repeated measurement is not the first step because the outcome is a risk-informed decision.

Expected:

Checks the catalog and routes to /ak:predict only if that installed skill owns the decision outcome; otherwise it reports the missing route instead of inventing an autoresearch command.

/ak:autoresearch "Expand edge-case coverage for the import parser; bound the file scope and produce test hypotheses before implementation" --metric coverage

Use when:

Use when the measurable result is better scenario coverage or test hypotheses rather than immediate code edits.

Expected:

Routes to /ak:scenario when available, treats coverage as the named metric, asks for feature or file scope, and lets the scenario workflow produce its own report rather than an implementation diff.

/ak:autoresearch "Run a threat-led review of webhook signature handling; keep the code scope authorized and mask any credentials in findings"

Use when:

Use when repeated verification should start by selecting a security-review route with explicit evidence boundaries.

Expected:

Discovers whether /ak:security is installed, routes authorized threat-led analysis there when it owns the outcome, and preserves the normal approval gate for external effects.