eve-bootstrap

Warn

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • Data Exposure & Exfiltration (MEDIUM): The skill accesses the sensitive ~/.ssh/ directory to list and read public keys for authentication requests. While public keys are non-secret, programmatic access to this directory is a high-risk pattern. The skill also communicates with a non-whitelisted external API (api.eh1.incept5.dev). Severity is adjusted to MEDIUM as these actions are core to the skill's primary function of developer onboarding.
  • Unverifiable Dependencies & Remote Code Execution (LOW): The skill directs the user to install the @anthropic/eve-cli package via npm. Per the [TRUST-SCOPE-RULE], this is classified as LOW severity because the package belongs to a recognized trusted organization.
  • Indirect Prompt Injection (LOW): The skill processes untrusted user input (e.g., org name, email, project slug) and interpolates it into shell commands and a configuration manifest (.eve/manifest.yaml). • Ingestion points: User input provided during the onboarding flow. • Boundary markers: Absent in command and file-write templates. • Capability inventory: Execution of eve CLI commands and file system write operations. • Sanitization: No explicit sanitization or validation of user-provided strings is performed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 19, 2026, 03:54 PM