Back to skill

Security audit

angel-agent-guard

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners report untrusted text being placed in a shell command and commands that download and run packages without a fixed version.

Consider the suspicious rating before installing. The results report shell-command and package-installation risks, as well as an unclear warning about sending supplied text to hosted checks. Skillspector reports complete coverage of the one listed component, SKILL.md; that coverage does not establish that the reported risks will be exploited or that the skill is safe.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md
Finding

Untrusted content is interpolated into a shell command

Content
View full analysis

Impact

Attacker-supplied commands can execute with the agent's filesystem, network and credential access before the security check runs.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

The default scanning workflow lacks a privacy warning about sending supplied text to a hosted service.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The skill directs the agent to scan any external text using a workflow with hosted checks, but does not clearly warn at that point that supplied content may leave the local environment or require consent. External text can still contain confidential correspondence, personal information, or proprietary material; mentioning hosted pricing and a separate offline command later does not provide an adequate data-handling boundary.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y angel-agent-guard'.

Medium
Category
MCP Rug Pull
Confidence
99% confidence
Finding

Although this is a CLI rather than an identified MCP server, the underlying supply-chain finding is valid: npx -y can download and execute an unpinned package without installation confirmation. A compromised release could execute with the agent's permissions, access scanned content and environment secrets, and potentially steal the payment key described elsewhere in the skill.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y angel-agent-guard'.

Medium
Category
MCP Rug Pull
Confidence
99% confidence
Finding

The purchase-verification command executes an unpinned npm package, allowing a compromised release to run with the agent's permissions. Its position immediately before payment increases the consequences: malicious code could falsify verification results or access credentials available to the process.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y angel-agent-guard'.

Medium
Category
MCP Rug Pull
Confidence
99% confidence
Finding

The POST-verification example repeats the unpinned package execution vulnerability; the --post flag does not mitigate it. Because this workflow concerns paid x402 endpoints, a compromised package could manipulate verification decisions or misuse payment credentials if they are available in its environment.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y angel-agent-guard'.

Medium
Category
MCP Rug Pull
Confidence
99% confidence
Finding

The token-check command downloads and executes a package without an exact version constraint. A compromised release could falsely label a malicious token as safe or influence the subsequent swap or approval workflow; the documented automatic-payment key would also be exposed if inherited by that process.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y github'.

Medium
Category
MCP Rug Pull
Confidence
99% confidence
Finding

The abbreviated static match refers to the full github:daoudandiaye98-alt/angel-agent-guard fallback, which has no immutable commit reference. This introduces a mutable source-code execution path outside the npm release path, so a repository compromise or unreviewed update could execute arbitrary code with the agent's permissions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y angel-agent-guard'.

Medium
Category
MCP Rug Pull
Confidence
98% confidence
Finding

This occurrence mentions the original unpinned npm command as the target of the fallback substitution, rather than adding a separate execution step. It nevertheless reinforces the same vulnerable installation guidance, and the alternative GitHub command does not establish an immutable or reviewed source.

Content

No source excerpt is available for this finding.