Agent Profiles - Cloudeval AI Docs

Agent Profiles change how Cloudeval organizes an evidence-grounded answer. They do not expand the evidence available to the selected project.

Available profiles

Profile id	Use it for	Expected emphasis
`architecture`	Topology, dependencies, blast radius, and Well-Architected review	Structure, affected resources, tradeoffs, and evidence gaps
`cost`	Cost drivers, savings, and FinOps decisions	Baseline, confidence, coverage, and prioritized savings
`triage`	Fast prioritization	Immediate risks, evidence gaps, and next checks
`remediation`	Sequencing fixes	Ordered actions, dependencies, validation, and rollback
`visual-explainer`	Visual walkthroughs	Graph-backed diagrams and missing-relationship caveats
`scripter`	Commands, YAML, scripts, and runbooks	Copyable output with dry-runs and rollback steps
`change-reviewer`	Pull requests and merge readiness	Current/base, proposed, live evidence, and PASS/WARN/BLOCK
`evidence-auditor`	Trust, freshness, and coverage	Provenance, timestamps, unsupported claims, and `Not assessed` areas
`security-reviewer`	Exposure, identity, secrets, and attack paths	Confirmed risks separated from gaps and hardening advice

Run a profile

cloudeval agents run change-reviewer \
  "Review this pull request" \
  --project <project-id>

Display names may contain spaces, but CLI and MCP automation use the profile ids shown above. Architecture includes the Well-Architected lens; there is no separate Well-Architected profile.

Last modified on September 1, 2026