troubleshooting-docs

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input from conversation history to generate documentation and search existing files.
  • Ingestion points: User-confirmed symptoms, exact error messages, and solution steps extracted in Step 2.
  • Boundary markers: No explicit delimiters or instructions are used to separate user data from the agent's instructions during documentation creation or searching.
  • Capability inventory: The skill uses Bash (grep, mkdir, echo, cat) and Write tools to interact with the file system.
  • Sanitization: While Step 4 mentions a sanitized symptom for filenames, no specific sanitization or escaping is mandated for error strings used in shell commands like grep.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (grep, mkdir, echo, cat) to manage the documentation library. This presents a risk of command injection if the agent interpolates unsanitized user input containing shell metacharacters (e.g., backticks, semicolons) into the command strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 01:25 PM