skills/base/skills/plugin-review/Gen Agent Trust Hub

plugin-review

Warn

Audited by Gen Agent Trust Hub on Oct 11, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill's live testing workflow in references/live-testing.md involves installing and running external code from the npm registry. It instructs the agent to use npm install and npx -y <pkg> --help to verify SDKs or CLIs mentioned in a plugin. Since the package name is derived from the untrusted Pull Request content, this creates a vector for executing arbitrary code from a malicious package during the review process.
  • [REMOTE_CODE_EXECUTION]: The use of npx to dynamically fetch and execute packages based on inputs found in a third-party Pull Request constitutes a remote code execution risk, as the agent may execute code from an attacker-controlled package.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from several sources, creating an attack surface for indirect prompt injection.
  • Ingestion points: The agent reads PR metadata, descriptions, and diffs using gh pr view and gh pr diff. It also fetches raw plugin files via gh api and interacts with external protocol APIs using curl as described in SKILL.md and references/live-testing.md.
  • Boundary markers: While the skill uses structured report templates (references/report-template.md), it lacks explicit boundary markers or instructions to the agent to disregard instructions embedded within the analyzed plugin markdown content.
  • Capability inventory: The agent has access to powerful CLI tools including gh (for commenting on PRs), curl (for network requests), and npm/npx (for code execution), which could be abused if an injection is successful.
  • Sanitization: No evidence of sanitization, escaping, or filtering of the ingested content is present in the skill's instructions.
  • [COMMAND_EXECUTION]: The workflow relies on executing system commands like gh, curl, and npx which are parameterized with data extracted from untrusted plugins, such as Pull Request numbers, package names, and API endpoints.
  • [EXTERNAL_DOWNLOADS]: The skill fetches the latest plugin specification from a trusted source on GitHub (https://raw.githubusercontent.com/base/skills/master/skills/wallet-mcp/references/plugin-spec.md). This is a legitimate operation for the skill's primary purpose.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 11, 2026, 06:48 AM