pr-feedback-classifier

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests and processes untrusted data from GitHub pull request comments and review feedback, which constitutes a surface for indirect prompt injection attacks. Although the output is constrained to a JSON schema, embedded instructions in comments could attempt to influence the agent's summaries or complexity assessments.\n
  • Ingestion points: External data is retrieved using the erk exec classify-pr-feedback command (SKILL.md).\n
  • Boundary markers: There are no explicit delimiters or instructions to treat the fetched content as untrusted or to ignore embedded directives.\n
  • Capability inventory: The skill identifies and returns thread and comment IDs used for subsequent automated actions like resolving threads or replying to discussions via the erk CLI.\n
  • Sanitization: No filtering or sanitization of the comment text is performed before processing.\n- [COMMAND_EXECUTION]: The skill uses the 'erk' CLI tool to execute commands for fetching and managing PR feedback. These operations are performed via shell execution based on provided arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 03:21 AM