Disclosed vulnerability
Agentjacking: injecting instructions into Sentry error events hijacks MCP-connected coding agents
AIR-2026-0136 AIR record · 2026-06-csa-agentjacking-sentry-mcp
From the supporting source
Attackers inject malicious instructions into Sentry error events using only a Sentry DSN-a public, write-only credential discoverable from browser JavaScript or GitHub search.labs.cloudsecurityalliance.org
The record
Disclosed
2026-06-12
Vendor
Sentry
Product
Sentry MCP server as consumed by Claude Code, Cursor and Codex
Risk domain
Agent behaviour › Unsafe actuation
Also touches
Data exfiltration channels, Confused deputy
System type
Generative model
Who acted
Agent acted — the MCP-connected agents acted on injected Sentry events
Attack surface
MCP server
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection error event
Impact type
Agent hijack, Unauthorized action, Data exfiltration
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI01 – Agent Goal Hijack
- MITRE ATLAS
- AML.T0048 – External Harms
- NIST AI RMF
- MEASURE 2.6; MANAGE 1.3
- EU AI Act
- Annex III §1(c)
- ISO/IEC
- TR 24028; 24029-1; 23894
- AIUC-1
- C. Safety
- Scope
- agent misuse
- Mitigation
- Require confirmation for destructive actions; Implement dry-run modes; Use layered safety controls.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
85% exploitation success across the three tested agents; the injection credential is a public write-only Sentry DSN
Capability abused
Defense & remediation
Recorded defense
Sentry content filter added after disclosure (payload-string blocklist only)
Guardrail kinds
Permission scope
Guardrail outcome
Silent
Fix shipped
Partial: Sentry declined root-cause remediation on 2026-06-03 and shipped only a reactive content filter
Taxonomy & references
Curator notes
CSA AI Safety Initiative research note. Sentry described the issue as 'technically not defensible' at the platform level. Quote's em-dash normalized to a hyphen. Also referenced in the wild as 'Tenet: Agentjacking'.
Supporting sources
Cite as
Enkrypt AI. "Agentjacking: injecting instructions into Sentry error events hijacks MCP-connected coding agents". AIR, AIR-2026-0136. https://agentincident.vercel.app/i/AIR-2026-0136/ (accessed 2026-10-19).