code-change-impact

Warn

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to automatically identify and execute shell commands for typechecking, building, and testing based on scripts found within project manifests like package.json, pyproject.toml, Makefile, and CI configuration files. This design allows for the execution of arbitrary code defined within the repository being analyzed.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface (Category 8) because it processes untrusted data from the repository environment to determine its verification logic and command execution.
  • Ingestion points: Repository manifest files and git diff output are ingested in SKILL.md Phase 1 and references/recipes.md Section 2 to identify project conventions and commands.
  • Boundary markers: The skill lacks explicit delimiters or instructions to prevent the agent from obeying malicious instructions or scripts embedded within the analyzed data.
  • Capability inventory: The skill uses shell access to run git, rg, and various project-specific tools (e.g., npm, pytest, mvn, cargo) discovered at runtime.
  • Sanitization: There is no evidence of command validation, sanitization, or a requirement for user approval before executing the discovered verification scripts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 15, 2026, 11:17 PM
Security Audit — agent-trust-hub — code-change-impact