{"resourceId":"futurum-cisco-project-quarry-agent-detection-analysis-20261030","versions":[{"version":"external-6613747638b73ac38fe29b9c3efaafd3ec6228af8634ac55ae7cc1083f52e454","resource":{"id":"futurum-cisco-project-quarry-agent-detection-analysis-20261030","title":"Futurum: Cisco Project Quarry brings detection-as-code to AI agents, but value depends on rule quality, full traces and neutral governance","organization":"The Futurum Group","sector":"Industry analyst firm (independent of Cisco)","geography":"Global","publishedAt":"Futurum Group, Oct. 30, 2026","publicationDate":"2026-10-30","eventDate":null,"sourceName":"Futurum Group","sourceLabel":"Cisco Wants Project Quarry to Be the Sigma of Agent Detection","sourceUrl":"https://futurumgroup.com/insights/cisco-wants-project-quarry-to-be-the-sigma-of-agent-detection/","evidenceClass":"independent-reporting","outcomeClass":"mixed","topics":["developers-agents","data-security","governance-procurement"],"finding":"Futurum describes Project Quarry (open-sourced Oct. 28) as a rule language and engine that evaluates whole agent sessions. It argues value is capped by rule-library quality, few SOC teams have tuned detections on agent traces, and Quarry looks like a Cisco-owned Snort-style format rather than a neutral Sigma-style standard. Its 1H2026 survey (N=929) found 55% lack technical ability to detect or contain a compromised AI agent. Futurum notes Cisco says Quarry does not replace runtime authorization.","sledRelevance":"Moderate for SLED agencies and universities piloting AI agents: provides an independent check on Cisco's agent-detection claims (archived Cisco blog, Oct. 28) and frames what to require before adopting.","evidence":"Analyst commentary plus a cross-industry survey; no testing of Quarry. Futurum discloses AI-assisted writing.","architectureImplications":"Requires full agent session traces (messages, tool calls, MCP) to be logged to be useful.","governanceImplications":"Watch for neutral governance; avoid committing to a single-vendor rule format without portability.","securityPrivacyImplications":"Detection does not replace runtime authorization for agent actions.","caveats":"Survey is not SLED-specific; no hands-on evaluation; analyst opinion; AI-assisted content disclosed.","streamIds":["cisco"],"roles":{"sales":"Sales framing: use this to set realistic expectations for agent security. Futurum supports the need, with 55% of surveyed decision-makers unable to detect a compromised agent, but says Quarry's value depends on rule quality and traces most SOCs lack. For SLED customers piloting agents, position a logging and governance readiness review before any detection tool. Do not present Quarry as a standard or as tested.","engineering":"Engineering framing: Quarry evaluates entire agent sessions, so its prerequisite is complete trace capture of messages, tool calls, tool results and MCP interactions. Check whether customer agent platforms emit such traces and whether they can reach Splunk. Test starter rules against recorded sessions to measure misses and false positives. Keep runtime authorization, such as least-privilege tool scopes, as a separate control, and note the format is Cisco-owned today.","delivery":"Delivery framing: a small pilot owned by the customer SOC with the AI platform team. Work is trace capture for one agent use case, running starter rules, and documenting tuning effort. Governance checkpoints include data retention of agent traces, which may contain sensitive records. Acceptance is proposed, such as full trace coverage for the pilot agent. Risks are trace gaps, rule maintenance burden and format lock-in."},"roleDetails":{"sales":{"customerProblem":"SLED teams deploying AI agents cannot yet detect agent misuse.","stakeholders":["CISO","SOC lead","AI program lead","Records/privacy officer"],"discoveryQuestions":["Do your agent platforms log full sessions including tool calls?","Who owns agent security monitoring?","Is runtime authorization in place for agent tools?"],"valueHypothesis":"Bounded: trace capture plus session rules may improve visibility into agent misuse.","potentialEngagement":"An agent-logging and detection readiness assessment for one use case.","unsupportedClaims":["That Quarry is an industry standard","Any detection accuracy","That detection replaces authorization"]},"engineering":{"fit":"Early-stage fit for SOCs already on Splunk piloting agents.","architectureIntegration":"Rules evaluate agent session traces; findings can flow to Splunk per Cisco.","prerequisites":["Full agent trace logging","Splunk or other SIEM ingestion","Rule tuning capacity"],"constraints":["Cisco-owned format; no neutral governance yet","Starter rules only","Traces may contain regulated data"],"security":"Pair with least-privilege tool scopes and approval for consequential actions.","proofOfValue":["Replay recorded sessions through starter rules","Measure false positives and misses","Confirm trace completeness"]},"delivery":{"work":"Trace capture, rule pilot and tuning record for one agent.","dependencies":["Agent platform logging","SIEM capacity","Privacy review"],"ownership":"Customer SOC with AI platform team.","skillsAdoption":"Analysts need agent-trace literacy.","governanceCheckpoints":["Privacy review of trace retention","Review of rule portability"],"acceptanceCriteria":["Proposed: 100% session trace coverage for pilot agent","Proposed: documented false-positive rate after tuning"],"risks":["Trace gaps","Rule maintenance burden","Lock-in"]}},"retrievedAt":"2026-11-03T04:05:35Z","enrichedAt":"2026-11-03T04:07:53Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Agent-trace detection requires new analyst skills.","procurementImplications":"Prefer portable formats; ask about governance roadmap in procurement.","operatingModelImplications":"Agent monitoring needs joint SOC and AI platform ownership.","sourceVerification":{"openedUrl":"https://futurumgroup.com/insights/cisco-wants-project-quarry-to-be-the-sigma-of-agent-detection/","referenceExcerpt":"The limit is familiar, in that the value is capped by the quality of the rule library.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}