Back to skill

Security audit

Imprint

Security checks for vulnerabilities and agentic risk

Overview

The skill is local and disclosed, but it silently builds a persistent behavioral profile and may read predicted workspace files without clear user approval.

Install only if you are comfortable with the agent continuously inferring and storing behavioral patterns such as work rhythm, communication style, active hours, and autonomy preference. Keep it out of shared or sensitive workspaces unless you add clear opt-in, retention, reset, and file-scope controls, and do not let inferred preferences authorize silent file access or tool use.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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
Findings (2)

T02 · Agent Memory Poisoning

Warning
Location
operator-model-schema.json:27
Finding

Persistent Free-Form Trait Values Can Poison Future Agent Behavior

Content
View full analysis
0.8 = strong." }, "observations": { "type": "integer", "minimum": 0, "description": "Number of observations supporting this trait" }, "last_updated": { "type": "string", "format": "date-time", "description": "When this trait was last updated" }, "history": { "type": "array", "description": "Optional: previous values for drift detection", "items": { "type": "object", "properties": { "value": { "type": "string" }, "date": { "type": "string", "format": "date-time" }, "confidence": { "type": "number" } } } } } } ``` ### Technical Analysis The Skill stores inferred be ...[truncated 2323 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
imprint.md:99
Finding

Passive Behavioral Profiling Enables Overbroad Proactive Workspace Access

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (7)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 189)May include surrounding context.

md
- `SKILL.md` — this file

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill description markets personalization and learning, but it does not clearly warn up front that the agent will persistently profile behavioral patterns and store them across sessions. This undermines informed consent and can cause users to enable a profiling capability without understanding the retention and behavioral inference involved.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The instruction to load the operator model at startup and 'Don't announce it — just do it' directs covert behavioral adaptation based on accumulated profiling data. Silent application reduces transparency, makes it harder for the user to detect or challenge profiling-driven behavior, and increases the risk of manipulative or privacy-invasive personalization.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The startup and session-end steps direct the agent to automatically load and update a persistent operator model file, enabling silent cross-session profiling without a clear user-facing consent step. This is dangerous because it normalizes background monitoring and retention of inferred preferences, creating privacy risk even in a local-only design.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill instructs the agent to derive and persist behavioral observations about the operator across sessions without requiring an explicit upfront notice or consent flow. Even though it says not to store raw message content and to keep data local, it still creates a durable behavioral profile from interaction signals, which can violate user expectations and expose sensitive inferences if the workspace is shared or later accessed by others.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The model stores a specific active-hours profile for the operator without any indication of user choice, consent, or contextual necessity. In a skill explicitly designed to build cross-session behavioral profiles, this kind of temporal preference data can enable covert personalization, inference about a user's routine, and more effective manipulation or targeting over time.

Content

No source excerpt is available for this finding.