Urgency and Pressure Manipulation in Agent Input

SIGMA detection rule by AgentShield · Unknown

Detects adversarial content that uses urgency, time pressure, or threat language to rush AI agent decision-making and bypass careful evaluation. These patterns exploit cognitive shortcuts analogous to social engineering techniques described in the "Semantic Manipulation Traps" category by Franklin et al. (2026).

title: Urgency and Pressure Manipulation in Agent Input
id: 63cb1a2f-a7e0-5d3d-9b0e-795d70849fbc
aliases:
  - agent-urgency-manipulation-001
status: test
description: |
  Detects adversarial content that uses urgency, time pressure, or threat
  language to rush AI agent decision-making and bypass careful evaluation.
  These patterns exploit cognitive shortcuts analogous to social engineering
  techniques described in the "Semantic Manipulation Traps" category by
  Franklin et al. (2026).
references:
  - https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6372438
  - https://attack.mitre.org/techniques/T1684/001/
  - https://owasp.org/www-project-top-10-for-large-language-model-applications/
author: AgentShield
date: "2026-04-06"
modified: "2026-07-03"
tags:
  - attack.initial-access
  - attack.t1566
  - attack.stealth
  - attack.t1656
logsource:
  product: ai_agent
  category: agent_events
detection:
  selection_urgency_keywords:
    event_type:
      - user_input
      - content_retrieval
      - tool_response
    message|contains:
      - 'do this immediately without'
      - 'critical deadline do not verify'
      - 'time-sensitive override'
      - 'emergency: skip all checks'
      - 'urgent: bypass approval'
      - 'no time to review just execute'
      - 'skip validation this is urgent'
      - 'execute now ask questions later'
  selection_threat_language:
    event_type:
      - user_input
      - content_retrieval
      - tool_response
    message|contains:
      - 'system will crash if you do not'
      - 'data loss imminent unless'
      - 'security breach in progress run'
      - 'production is down execute'
      - 'failure to comply will result in'
      - 'you will be shut down if'
      - 'your access will be revoked unless'
  selection_skip_safety:
    event_type:
      - user_input
      - content_retrieval
      - tool_response
    message|contains:
      - 'skip safety checks'
      - 'bypass security for this one'
      - 'disable verification temporarily'
      - 'override safety protocols'
      - 'ignore warnings and proceed'
      - 'suppress all alerts and run'
  condition: selection_urgency_keywords or selection_threat_language or selection_skip_safety
falsepositives:
  - Legitimate incident response instructions
  - Production outage remediation scripts
  - Time-critical but authorised deployment procedures
level: high

[[ currentRule.title && currentRule.title.length > 60 ? currentRule.title.slice(0, 60) + '…' : currentRule.title ]]

1cfbf96b-d390-4944-a87f-bcaa418c0e5f

Propose edits, view related rules, and engage with the community through comments.

Executable code, not a declarative rule

This submission is arbitrary [[ currentRule.format ? currentRule.format.toUpperCase() : '' ]] code, not a declarative detection rule — review it as you would any script before running it, and note that it cannot be safely aggregated or auto-executed alongside YARA/Sigma/Suricata/CRS rules.

Rejected pattern present

[[ reason ]]

What does this mean?

Format & documentation suggestions

Not a security issue — this rule is valid and was accepted as-is. These are optional fields/conventions its own format recommends, mainly affecting its quality score and how well it documents itself for other reviewers.

[[ reason ]]

What do these messages mean?

[[ currentRule.title && currentRule.title.length > 60 ? currentRule.title.slice(0, 60) + '…' : currentRule.title ]]

[[ currentRule.title ]]

Community Discussion