Compaction Laundering Attack on Context-Compacting Coding Agents
Researchers demonstrate an exploitable trust boundary in popular open-source coding agents where context compaction destroys message provenance. Attackers can forge user turns that survive conversation summarization and persist in the agent's memory.
- first report
- 22 Oct 2026
- last update
- 22 Oct 2026
Timeline · 1
Sources · 1
Research and catalogues
- 2026-10-22Compaction Laundering: Provenance Destruction in Context-Compacting Coding AgentsarXiv · arxiv.org/abs/2610.29869
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Research
- AI role
- Attack on AI
- categories
- Agent hijackPrompt injection
- severity
- 1 · Research / proof of concept
- targets
- Framework / tooling, Enterprise deployment
- owasp llm top 10
- LLM01:2025 Prompt InjectionLLM06:2025 Excessive Agency
- dates
- happened 22 Oct 2026 · first reported 22 Oct 2026
- origin
- News & research feeds
- record
- confidence 0.95 · created 2026-10-22 04:45Z · updated 2026-11-06 00:23Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# Compaction Laundering Attack on Context-Compacting Coding Agents https://gcdradar.vercel.app/incidents/compaction-laundering-attack-on-context-compacting-coding-ag · AIR-2026-1462 · closed 2026-11-05 - Lane: Research - Severity: 1 (Research / proof of concept) - Fix status: Fix status unknown - First reported: 2026-10-22 - AI role: Attack on AI ## Summary Researchers demonstrate an exploitable trust boundary in popular open-source coding agents where context compaction destroys message provenance. Attackers can forge user turns that survive conversation summarization and persist in the agent's memory. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2026-10-22: Compaction Laundering: Provenance Destruction in Context-Compacting Coding Agents (https://arxiv.org/abs/2610.29869) ## Sources - 2026-10-22 arXiv: Compaction Laundering: Provenance Destruction in Context-Compacting Coding Agents https://arxiv.org/abs/2610.29869 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- liveMobileProbe: Adaptive Security Evaluation of Mobile Agents via Trajectory-Aware Injection
Research · since 13 Nov 2026
AI - livePrompt Injection, Esq.: pwning AI agent guardrails with lawyerspeak
Research · since 11 Nov 2026
AI - liveLLM agents Defending LLM Agents Against Parameter Poisoning
Research · since 4 Nov 2026
AI - liveLLM search agents Adversarial Self-Evolving Attacks on Search Agents
Research · since 4 Nov 2026
AI - liveLLM agents Weaponized Safety Warnings for Controlling AI Agents
Research · since 4 Nov 2026
AI - closedCoding Agents CARI: Cross-Boundary Adversarial Runtime Injection in Coding-Agent Workflows
Research · since 30 Oct 2026
AI