Back to skill

Security audit

Agent Cognitive Os

Security checks for vulnerabilities and agentic risk

Overview

This is a broad planning and verification skill for agent behavior, with no code payload or hidden install behavior found.

Install this if you want broad agent planning, orchestration, and verification behavior. Be aware that much of the guidance is in Indonesian and that the skill is intended to activate for high-level agent tasks, so users who prefer narrow task-specific behavior should review the scope before enabling it.

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)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The usage instruction says to use the skill 'when the conditions in the description are met' but does not define concrete activation boundaries or exclusions. For a top-level control/orchestration skill, this ambiguity can cause over-broad invocation, unnecessary privilege over planning and plugin coordination, and unsafe delegation decisions in downstream agent behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill content is predominantly written in Indonesian and directly instructs behavior and outputs without checking the user's preferred language. This can cause misunderstandings of constraints, confirmations, or risk prompts, which is especially problematic for a control-layer skill governing autonomous execution and verification.

Content

No source excerpt is available for this finding.

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · SKILL.md (reported line 935)May include surrounding context.

md
NEVER FABRICATE
NEVER CLAIM UNVERIFIED SUCCESS
NEVER RETRY FOREVER
NEVER IGNORE CURRENT STATE
NEVER CONFUSE ASSUMPTION WITH FACT
NEVER USE EXCESSIVE PRIVILEGE

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 1051)May include surrounding context.

md
| Acting without a plan | Hierarchical planning first |
| Losing state between steps | Track state explicitly |
| Ignoring plugin orchestration | Coordinate multi-plugin flows |
| No verification gate | Verify each milestone |

## Red Flags

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly says to activate for 'any top-level agent task' and describes very broad applicability, which can cause it to override narrower, task-specific safeguards or reasoning modes. In context, this is an agent-control skill that influences planning, tool use, plugins, and completion logic, so overly broad activation increases the chance of inappropriate autonomous orchestration on ordinary tasks.

Content

No source excerpt is available for this finding.