onboard

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill operates locally to establish a context file (~/.claude/company-context.md) based on user-provided information. This behavior is consistent with the skill's stated purpose and does not involve any external network requests or sensitive system access.
  • [COMMAND_EXECUTION]: The skill documentation includes a bash command (ln -sf) to create a symbolic link for file persistence. This is a standard file management operation for local workflows and does not require administrative privileges.
  • [PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by ingesting and storing founder responses as authoritative context for other agents.
  • Ingestion points: User-provided interview data stored in ~/.claude/company-context.md.
  • Boundary markers: Absent; the file uses markdown headers but does not implement delimiters to distinguish user content from instructions.
  • Capability inventory: File writing to the local user directory and documentation of local shell commands.
  • Sanitization: Absent; instructions encourage capturing verbatim responses from the user.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 01:55 PM
Security Audit — agent-trust-hub — onboard