LLM Testing: Systematically validate factual accuracy, refusal behavior, and safety boundaries
Build LLM tests for factual accuracy, citations, refusal, privacy, latency and cost; pin model settings and repeat runs to assess stochastic output.
#Agent Skills
AI-Assisted Testing: Improve testing efficiency while retaining test evidence
Assess AI-generated test ideas against real interfaces, missing boundaries, and traceable evidence before adding them to a test suite.
Prompt Testing: Systematically validate prompt accuracy, boundaries, and multi-turn consistency
Evaluate prompt changes across varied inputs and turns with pinned model settings, rubric-based scoring, boundary cases and version-regression checks.
Security Testing: Systematically identify security risk from assets, permissions, and attack surfaces
Prioritize security risks across identity, input, data, and dependency boundaries by exploitability and impact; record reproduction, fixes, and retests.
Multi-Role Quality Synthesis: Preserve different perspectives and reach actionable quality decisions
Synthesize product, QA, engineering and operations views while preserving evidence conflicts, assumptions and unresolved risks for release decisions.
Project Delivery Perspective: Include scope, dependencies, and delivery risk in quality decisions
Assess delivery readiness from scope, dependencies, milestones and evidence; make trade-offs and owners clear when plans shift.
Technical Quality Perspective: Include architecture, reliability, and maintainability in quality decisions
Connect architecture, availability, maintainability, recovery and dependencies to operational outcomes for a grounded system quality review.
UX Quality Perspective: Bring user task flows and usability issues into quality decisions
Assess task flow, waiting, feedback, accessibility and recovery to find experience risks even when the interface reports no errors.
Product Quality Perspective: Bring user value and business impact into quality decisions
Evaluate product quality through user tasks, business rules, trust and conversion risk so release decisions reflect customer outcomes.
QA Quality Perspective: Use coverage, risk, and evidence to support quality conclusions
Use coverage, execution evidence, untested paths and residual risk to support release decisions without overstating green test results.
Production Incident Analysis: From incident timeline to containment, validation, and retrospective actions
Stabilize production impact, build an evidence-backed timeline, and test competing hypotheses before naming an incident root cause.
Production Verification: Verify business flows, metrics, logs, and rollback conditions after release
Verify releases against real business flows, metrics, logs, and alerts. Set an observation window and criteria for continuing, pausing, or rolling back.