paragon-try-tools

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements secure local credential management by instructing the user to store sensitive signing keys and environment configuration in a dedicated hidden directory (~/.paragon) with restricted file permissions (chmod 600 and 700), which is a standard security practice for development tools.
  • [SAFE]: Authentication tokens (JWTs) are generated locally using a provided shell script (scripts/paragon-env.sh) that utilizes the system's openssl binary. The script is transparent, runs entirely on the local machine, and is designed to avoid exposing secrets to the console or agent logs.
  • [SAFE]: Network operations are scoped to official vendor infrastructure (agentos.useparagon.com and zeus.useparagon.com). The skill uses curl for these interactions, and the vendor infrastructure matches the author identity (useparagon).
  • [SAFE]: The instructions include a robust safety guardrail requiring the agent to wait for explicit user confirmation before executing any tools that have potential side effects (write, delete, or send operations).
  • [SAFE]: While the skill ingests data from external tool outputs (Category 8 surface), it includes specific instructions to "Report the outcome plainly," which guides the agent to treat the output as data rather than instructions, mitigating the risk of indirect prompt injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 08:16 PM