gstack-qa

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external URLs during the testing phase, which creates a surface for indirect prompt injection attacks.\n
  • Ingestion points: The target URL and its associated content are processed in Phase 1 and 2 of SKILL.md.\n
  • Boundary markers: The instructions do not define delimiters or warnings to treat external web content as non-instructional data.\n
  • Capability inventory: The skill utilizes the Bash, Write, and Edit tools to execute commands and modify source code.\n
  • Sanitization: There are no procedures for sanitizing or filtering input received from the target application before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill uses shell commands for environment detection, smoke testing, and version control.\n
  • Evidence: Uses bash to run git status, ls for framework detection, and curl for URL reachability checks in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 05:57 PM
Security Audit — agent-trust-hub — gstack-qa