pessimistic-code-review

Pass

Audited by Gen Agent Trust Hub on Apr 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes project-specific test commands (e.g., pytest, npm test, php artisan test, go test) to independently verify code functionality. These commands are executed in the local environment as part of the Phase 3 verification process.
  • [PROMPT_INJECTION]: The skill contains a behavioral override instruction requiring the agent to always respond in Italian regardless of the user's language setting, which restricts standard agent interaction behavior.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted inputs—such as pull request descriptions, task specifications, and code diffs—while possessing high-privilege capabilities like file writing and command execution.
  • Ingestion points: Untrusted data enters the context via the task specification (PR body, ticket) and the code files or diffs being reviewed (SKILL.md).
  • Boundary markers: No explicit delimiters or markers are used to isolate untrusted input content from the agent's logic, increasing the risk of the agent following embedded instructions.
  • Capability inventory: The agent can execute shell commands for testing and modify files to apply fixes (SKILL.md, Steps 5 and 6).
  • Sanitization: There is no specified logic for sanitizing or escaping external content before it is interpolated into prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 3, 2026, 12:31 PM