Back to skill

Security audit

evolver-solidify

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners report instructions to evade command safety checks, work around Hub review, and publish local assets without a privacy warning; scan coverage was partial.

Consider avoiding installation until the safety-policy and Hub review workarounds are removed or clarified, and the exact data intended for publication can be reviewed for privacy. The scanners inspected 4 of 5 components; tool-misuse analysis was degraded, so the evidence is incomplete. The suspicious verdict indicates reported security weaknesses, not a conclusion that the author is malicious.

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md
Finding

Explicit instructions to evade command safety enforcement through WSL2

Content
View full analysis

Impact

Following these instructions can permit execution of validation commands the original environment would reject. No particular malicious validation payload or successful runtime bypass is established by the available files.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code is only a metadata export and does not implement the substantive workflow described. It also shows no resource access, triggers, or undeclared capabilities. This assessment is limited to the supplied code chunk.

Content

No source excerpt is available for this finding.

Presents moving execution to WSL2 as a workaround for a security-policy rejection.

High
Category
Prompt Injection
Confidence
97% confidence
Finding

The skill explicitly recommends changing execution environments to avoid a safety-policy rejection, rather than resolving the rejected command safely. Because this workflow executes generated plans and validation commands, bypassing enforcement could allow unsafe commands to run, although the file does not demonstrate a specific malicious payload.

Content

No source excerpt is available for this finding.

The publish workflow lacks a privacy warning before sending local evolution assets to an external Hub.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

Publishing to an external Hub is part of the stated purpose, so external transmission alone is not suspicious. However, the instructions recommend manually uploading local asset data after auto-publication is blocked, without requiring privacy review or distinguishing a transport failure from a sensitive-data rejection; the nearby LeakCheck discussion makes this omission materially risky.

Content

No source excerpt is available for this finding.

Recommends manual publication as a workaround for a Hub review-policy rejection.

Medium
Category
Prompt Injection
Confidence
88% confidence
Finding

The blocker table treats a Hub review-policy rejection as something to work around through manual publication instead of resolving the review requirement. This encourages use of an alternate submission path without establishing that equivalent approval controls apply, though there is no evidence that the web UI actually bypasses server-side enforcement.

Content

No source excerpt is available for this finding.