Reusable AI capabilities
Skills are reusable instructions for a focused capability. Each Skill is based on a published Amo.ng Prompt or Workflow and explains what information you need, how to apply the method, and what completion looks like. Skills do not run automatically.
Create a defensible evidence-screening and extraction protocol for systematic, scoping, or living reviews, including eligibility rules, deduplication, reviewer decisions, appraisal, and flow reporting.
Assess evidence for paid-media creative fatigue against audience saturation, delivery changes and other confounders, then propose controlled refresh tests without treating descriptive patterns as causal proof.
Diagnose intermittent CI or local test failures, identify the supported nondeterministic cause, and define a cause-specific stabilization plan with reliability evidence instead of hiding product defects.
Turn a supplied source corpus into a literature synthesis that preserves study-level evidence, quality concerns, agreements, contradictions, uncertainty, and open research questions.
Review Google Tag Manager, GA4, ads, and consent-mode evidence to identify missing, duplicate, or unreliable conversion events before a tracking change is published or trusted for reporting.
Find where untrusted runtime inputs can bypass TypeScript’s compile-time assumptions, then specify focused validation, compatible error handling, and regression checks for affected consumers.
Develop method-compatible qualitative coding records from traceable excerpts, preserving rival interpretations, reflexivity, privacy and the appropriate human calibration or analytic-dialogue process.
Analyze lifecycle email deliverability evidence to identify sender, authentication, consent, audience, reputation, content, suppression, and provider issues, then produce a staged recovery plan with guarded sending controls.
Review GitHub Actions workflows, secrets, token scopes, runner trust boundaries, and repository protections to identify credential exposure and permission hardening priorities before changing CI/CD security controls.
Evaluate whether a consequential analytical conclusion remains credible under plausible changes to data quality, cohort definitions, metric logic, assumptions, missing-data treatment, model choices, and sensitivity scenarios.
Design a governed AI support triage pilot with knowledge-readiness checks, ticket classification rules, safe drafting limits, escalation paths, sensitive-data controls, human review gates, monitoring, and pilot success criteria.
Assess whether a FastAPI service is ready for a controlled production release by reviewing API boundaries, security, deployment topology, workers, observability, operational ownership, rollback readiness, and unresolved evidence gaps.
