open-gstack-browser

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the 'bun' installer from 'https://bun.sh/install' if the runtime is missing. This source is a well-known service, and the skill performs a SHA-256 checksum verification before execution.\n- [REMOTE_CODE_EXECUTION]: The script executes the downloaded 'bun' installer via bash. This execution is guarded by checksum validation and targets a well-known service.\n- [COMMAND_EXECUTION]: The skill interacts with several local binaries provided by the vendor (located in ~/.claude/skills/gstack/bin/) for configuration, telemetry, and repo management. It uses eval and source to process the output of these binaries.\n- [DATA_EXFILTRATION]: Telemetry regarding skill usage is collected and stored in ~/.gstack/analytics/. Users are prompted to configure telemetry preferences (including opting out) during the skill's initial run.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its browser automation capabilities.\n
  • Ingestion points: Web page content is ingested via the $B snapshot command in SKILL.md.\n
  • Boundary markers: No explicit delimiters or instructions are used to separate external web data from agent instructions.\n
  • Capability inventory: The skill can execute shell commands via Bash, read files, and prompt the user.\n
  • Sanitization: External content is not sanitized before being processed by the agent. This is a common and expected risk for browser-focused skills.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 02:25 PM