Back to skill

Security audit

Self Apply Pressure

Security checks for vulnerabilities and agentic risk

Overview

This skill is transparent and contains no code, but it broadly pressures an agent to act first and ask later without clear safety boundaries.

Install only if you want a more forceful troubleshooting style. Keep higher-priority instructions requiring confirmation before destructive changes, credential use, network-affecting actions, public posting, purchases, or broad system modifications.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: self-apply-pressure
description: Prevents AI from slacking off, getting stuck, shifting blame, asking empty questions, or declaring completion without verification. Core objective: exhaust all options, proactively troubleshoot, and deliver with evidence.
---

# Self-Applied Pressure

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 8)May include surrounding context.

md
---
name: self-apply-pressure
description: Prevents AI from slacking off, getting stuck, shifting blame, asking empty questions, or declaring completion without verification. Core objective: exhaust all options, proactively troubleshoot, and deliver with evidence.
---

# Self-Applied Pressure

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly tells the agent to 'do first, ask later' and prioritize reading files and running commands before asking questions. In a real environment, this can lead to unauthorized or harmful actions, especially when commands may modify systems, access sensitive data, or cause side effects without informed user approval.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger conditions for entering 'high-pressure mode' are broad and subjective, such as being 'tempted' to say something or perceiving insufficient initiative. In an agent skill, this can push the model toward over-persistent autonomous behavior, reducing opportunities to stop, seek clarification, or apply appropriate safety constraints before taking actions.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

This rule treats hesitation, manual handoff, or attributing issues to the environment as reasons to escalate autonomous troubleshooting. In context with the rest of the skill, it pressures the agent to continue acting instead of pausing for confirmation, which can override prudent restraint and lead to unsafe or unauthorized operations.

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

md
- Repeatedly fine-tuning old solutions without changing direction
- Tempted to say "I cannot resolve this"
- Suggesting the user handle something manually
- Attributing issues to the environment without verification
- Drawing conclusions without searching, reading source code, or consulting documentation
- Not verifying after a fix
- User explicitly requests "try again" or "try a different approach"