scoped-grill-me

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses file system operations to create and update a session ledger in the system's temporary directory. These operations are limited to data persistence and do not involve executing arbitrary code or shell commands.
  • Evidence: The agent is instructed to use 'platform-native' facilities to resolve the OS temporary directory and create a uniquely named Markdown file to store the session 'ledger'.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it persists user-provided information into its ledger, which is subsequently re-read and displayed to influence the agent's behavior.
  • Ingestion points: User answers that explicitly refine or change the goal or scope are recorded in the ledger (SKILL.md).
  • Boundary markers: The session data is structured using Markdown headers such as ## TL;DR and ## IN SCOPE (SKILL.md).
  • Capability inventory: The agent has permissions to read and write to the temporary directory.
  • Sanitization: No explicit sanitization or filtering of user-provided content is described.
  • [SAFE]: The skill follows security best practices for session data and lacks any mechanisms for data exfiltration.
  • Evidence: No network-related commands (such as curl or wget) or external URLs are present in the skill.
  • Evidence: The instructions explicitly state 'Do not hardcode a platform-specific path or write the ledger into the repository,' ensuring data is confined to temporary, session-specific locations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 06:00 PM
Security Audit — agent-trust-hub — scoped-grill-me