Case jqwik-agent-delete-output-instructionPrompt injectionWorldwide
Undisclosed jqwik addition instructs AI coding agents to delete app output
An undisclosed addition to the jqwik library contained instructions aimed at AI coding agents, telling them to delete application output, according to Ars Technica.
MediumAttack on AIStatus: Under investigationVerified case file
Open in the interactive mapCite this case
- Reported
- Oct 15, 2026
- Location
- Worldwide
- Target
- Developers running AI coding agents on jqwik
- Severity
- Medium
- Triggered by
- Unknown
- Actor type
- Unknown
- Reported loss
- None reported
- CVE
- None
What happened
An undisclosed addition to the jqwik library contained instructions aimed at AI coding agents, telling them to delete application output, according to Ars Technica.
Impact: Agents could be steered into deleting application output. The number of affected users is not stated.
Who triggered it
Actor: Unknown
Attribution: The summary does not say who added the instructions.
Motive: Unknown
How it happened
Attack vector: Undisclosed addition to the jqwik project containing instructions that AI coding agents read and may follow.
Root cause: Agent-directed instructions embedded in a dependency, which coding agents may treat as trusted input.
Timeline
- Reported by arstechnica.com via Hacker News
Response
Unknown
Current status: Under investigation
Sources (open in a new tab)
- arstechnica.com via Hacker News arstechnica.com
Cite this case
Undisclosed jqwik addition instructs AI coding agents to delete app output. AI Threat Watch, case jqwik-agent-delete-output-instruction, Oct 15, 2026. https://aithreat.io/cases/jqwik-agent-delete-output-instruction/
Permanent link: https://aithreat.io/cases/jqwik-agent-delete-output-instruction/