find-companies-using-specific-tech-stack
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands to resolve the local directory path and execute helper scripts (
api.shandsignup.sh). It also instructs the agent to interpolate user-provided values, such as email addresses and OTP codes, into these commands. While this is a standard pattern for interactive skills, it relies on the agent correctly quoting input to prevent shell injection.\n- [CREDENTIALS_UNSAFE]: The skill implements a secure credential management system by storing API tokens in a local.envfile and automatically updating.gitignoreto prevent accidental disclosure. This is a best practice that isolates sensitive keys from the AI's direct conversation history and platform logs.\n- [DATA_EXFILTRATION]: Network activity is restricted to the vendor's domain (api.openfunnel.dev), which is required for the skill's documented functionality of performing technographic lookups and user authentication.\n- [PROMPT_INJECTION]: The skill processes job posting data from an external API, creating a surface for indirect prompt injection.\n - Ingestion points: Technographic data and job post summaries returned by
api.sh.\n - Boundary markers: Absent in the markdown formatting of results.\n
- Capability inventory: Subprocess execution via local bash scripts.\n
- Sanitization: No explicit sanitization of the external API content before presentation.
Audit Metadata