Back to skill

Security audit

End-User Software Training

Security checks for vulnerabilities and agentic risk

Overview

The skill’s training workflow is coherent, but its update path can run server-provided commands and its installer uses weakened publisher-signature checks, so it should be reviewed before installation.

Install only if you trust PersonWise with the selected training materials and account actions. Review any update command carefully before approving it, because the skill instructs the agent to run update commands returned by the service. Be aware the installed CLI persists in your user profile and the current package does not enforce native publisher-signature verification on macOS or Windows.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (3)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:113
Finding

Server-Controlled Update Commands Can Be Executed Locally

Content
View full analysis
` with the directory of this installed Skill (the directory containing this Skill's SKILL.md). Never run `doctor` or a generic capability preflight to check freshness; the `update check` command above is the freshness check. Never ask more than once per component per session, and never substitute another command, flag, origin, or download path for the printed `action`. ``` ### Technical Analysis The Skill treats an `action` command returned by the remote PersonWise service as executable shell text. It instructs the Agent to run that command exactly after obtaining general user approval, but does not require local validation of: - The executable being invoked - Permitted command-line arguments - The download origin - The installation destination - Shell operators, redirections, substitutions, or addi ...[truncated 1891 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
assets/bootstrap.sh:29
Finding

Unix and macOS Bootstrap Permits Installation Without Native Signature Verification

Content
View full analysis
&2 exit 9 fi ``` The later native-signature check is also conditional on a status that is never `verified` for the included macOS artifacts: ```sh if [ "$(uname -s)" = "Darwin" ] && [ "$native_signature_status" = "verified" ]; then codesign --verify --deep --strict --verbose=2 "$candidate" spctl --assess --type execute --verbose=2 "$candidate" actual_identity="$(codesign -dv --verbose=4 "$candidate" 2>&1 | sed -n 's/^Authority=//p' | head -1)" if [ -z "$native_identity" ] || [ "$actual_identity ...[truncated 2393 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
assets/bootstrap.ps1:10
Finding

Windows Bootstrap Installs an Executable with Authenticode Verification Disabled

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

Scope Creep

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The bootstrap script performs an external download and installs executable software even though the skill metadata declares only Bash permission and the skill’s primary purpose is end-user software training. Although the script includes multiple integrity checks and explicit approval gating, it still expands the skill’s effective capabilities into software acquisition and installation, which is dangerous if the permission model or user expectations rely on the manifest being complete and accurate.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This skill is described as providing grounded end-user software training, but the included bootstrap logic installs, upgrades, and rolls back a local executable. That mismatch increases the risk of confused-deputy behavior: a user invoking a training skill may not reasonably expect local software modification, making consent, review, and policy enforcement less reliable even if the installer itself is not overtly malicious.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · assets/bootstrap.sh (reported line 165)May include surrounding context.

sh
echo "Release executable checksum mismatch." >&2
  exit 6
fi
chmod 700 "$candidate"
if [ "$(uname -s)" = "Darwin" ] && [ "$native_signature_status" = "verified" ]; then
  codesign --verify --deep --strict --verbose=2 "$candidate"
  spctl --assess --type execute --verbose=2 "$candidate"