electron-chromium-upgrade

Pass

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies failed build targets and patch conflicts from command output to automate subsequent actions. This ingestion of untrusted external data (build/sync logs) could potentially be exploited by an attacker who can influence build errors to inject malicious instructions. \n
  • Ingestion points: Error output from e sync --3 and e build -k 999 (SKILL.md).\n
  • Boundary markers: None identified; the agent parses raw command output directly.\n
  • Capability inventory: Execution of shell commands via e and git, local file system modification, and deletion of cache directories (SKILL.md).\n
  • Sanitization: None; the skill does not specify validation or escaping of parsed target names or error messages before using them in further commands.\n- [COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands, including git am, git rebase, and directory cleanup. While standard for the task, the dynamic construction of commands like e build -t {target}.o from build log output presents a potential command injection path if the input is not strictly validated.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 6, 2026, 04:52 PM
Security Audit — agent-trust-hub — electron-chromium-upgrade