write-bug-report

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill directs the agent to retrieve and execute commands found in the repository's AGENTS.md file to reproduce defects. This pattern relies on the integrity of the project's documentation to avoid arbitrary command execution.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the codebase and external documentation.
  • Ingestion points: The agent reads command strings from AGENTS.md, searches the codebase using grep, and references requirement IDs from external specifications.
  • Boundary markers: No specific delimiters or safety warnings are provided to prevent the agent from obeying malicious instructions embedded in the project files or commands.
  • Capability inventory: The agent is authorized to execute shell commands (reproduction) and write files to the local file system (artifact creation).
  • Sanitization: The instructions do not include validation steps for the commands extracted from AGENTS.md before execution.
  • [DATA_EXFILTRATION]: The instructions require the agent to resolve the home directory (~) and write diagnostic artifacts to ~/.agents/artifacts/. While this is a standard developer workflow, it confirms the agent's capability to access and write to locations outside the repository root.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 06:02 PM
Security Audit — agent-trust-hub — write-bug-report