Testing an assembled agent — loop, tools, guardrails, the whole system — before you ship a change: golden scenarios from real traces, assertion layers from final answers down to individual tool calls, the mock-and-stub decision for each tool, replay as regression testing, flaky-agent triage, environment simulation that stays honest, and the written harness record. It builds the pre-release test layer for agent systems; it never predicts agent performance and never certifies an agent as safe. Use when asking "test an agent", "agent eval harness", "golden scenarios", "tool call assertions", "agent regression suite", "mock tool", "agent test flakiness", "pre-ship agent test", "LLM agent CI", "assertions on trajectory". Not for model selection suites (ac-model-evaluation), not for live-run inspection and dashboards (ac-agent-run-observability), not for the runtime interrupt layer (ac-agent-guardrails), not for unit-testing software in general (ac-testing).
cp -r cryptoskill/skills/exchanges/anoscranel-ac-agent-eval-harness .claude/skills/clawhub install anoscranel-ac-agent-eval-harnessCapabilities below are detected automatically by an open-source scanner that reads the skill's text and scripts (see how this is computed). Not measured means the scanner couldn't make a confident call — it is NOT a green check, and you should treat it as a possible red flag until a human or a stronger scanner has measured it.
read_only Phase 1 single-mode classification — multi-mode breakdown deferred to Phase 2.
We did not find any well-known hosted services in this skill's text or scripts. This does NOT mean the skill is local-only — it might use services we don't yet recognize, or talk to them through code paths our scanner can't reach. A complete dependency list (every package, library, and binary, with integrity hashes) is on the roadmap; today we only show recognized hosts.
No one has audited this skill yet. That is different from “audited and clean” — it just means no professional reviewer (a security firm, the CryptoSkill team, or a verified independent researcher) has signed off on it. There are no audit reports to read. How reviewer levels work →
SKILL.md → SOURCE.md → TRUST.auto.yaml → Browse directory →
Auto-generated by cryptoskill/extract-capabilities/0.3.1 · hosted-service list version 2026-09-06 · how this is computed