skills/open-gsd/gsd-core/gsd-fast/Gen Agent Trust Hub

gsd-fast

Pass

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes natural language task descriptions provided via the argument-hint. If these tasks involve processing external or untrusted data (e.g., refactoring a file containing malicious comments), the agent is susceptible to indirect prompt injection. There are no explicit boundary markers or instructions to disregard embedded commands in the input data.
  • [COMMAND_EXECUTION]: The skill explicitly allows the use of the Bash tool to perform trivial tasks. This grants the agent the ability to execute arbitrary shell commands. While restricted by the agent's own planning logic, the lack of specific command whitelisting means the capability surface is broad.
  • [DATA_EXFILTRATION]: The skill provides a capability chain consisting of file discovery (Glob), file reading (Read, Grep), and file writing or command execution (Write, Edit, Bash). This combination provides the technical means to access sensitive local data and potentially exfiltrate it if the agent is successfully manipulated.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 11, 2026, 09:38 PM
Security Audit — agent-trust-hub — gsd-fast