ralph-specum-feedback

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted external data such as logs, environment details, and project files.
  • Ingestion points: The skill reads affected files, commands, and log files as specified in the Action section of SKILL.md.
  • Boundary markers: There are no explicit boundary markers or instructions to the agent to treat the ingested data as untrusted or to ignore instructions embedded within the logs/files.
  • Capability inventory: The skill uses the gh command-line tool for issue creation, representing a command execution capability.
  • Sanitization: No sanitization or validation of the ingested content is described before it is processed or submitted.
  • [COMMAND_EXECUTION]: The skill uses the gh (GitHub CLI) tool to create issues. This involves executing shell commands to interact with a well-known external service (GitHub), which is consistent with the skill's primary purpose of submitting feedback.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:36 PM
Security Audit — agent-trust-hub — ralph-specum-feedback