skill-updater

Fail

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/main.cjs file contains a function updateRoutingTableKeywords that performs direct string interpolation of the skill name into a CommonJS configuration file (routing-table-intent-keywords-data.cjs). The code uses '${name}': within a template literal without sanitizing the name input for control characters like single quotes. A craftily named skill (e.g., containing ' and JavaScript commands) could inject and execute arbitrary code when the routing table is subsequently loaded by the project's runtime.
  • [REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md direct the agent to fetch content from the external repository VoltAgent/awesome-agent-skills and decode it using base64 -d via the Bash tool. While the output is intended for reading and comparison, this pattern of downloading and decoding content from an untrusted third-party source into the local execution environment represents a significant supply chain risk.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external skill definitions and research papers. Although it includes a "Security Review Gate" for the agent to manually scan for malicious patterns, this mitigation is bypassable. Sophisticated prompt injections or hidden instructions in the fetched content could influence the agent's behavior, leading it to perform unsafe file modifications or bypass safety protocols during the update workflow.
  • [EXTERNAL_DOWNLOADS]: The skill performs automated network operations to fetch data from github.com/VoltAgent/awesome-agent-skills and various other GitHub repositories discovered during research. These sources are not recognized as trusted organizations or well-known services, increasing the risk of ingesting malicious or unverified content into the development environment.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 16, 2026, 02:56 AM