plannotator-annotate

Warn

Audited by Gen Agent Trust Hub on May 9, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute plannotator annotate <path-or-url> using a Bash shell. If the <path-or-url> argument is derived from untrusted user input without sanitization, an attacker could execute arbitrary commands by injecting shell metacharacters such as semicolons, pipes, or backticks.
  • [PROMPT_INJECTION]: The skill processes content from external files and URLs, exposing the agent to indirect prompt injection vulnerabilities.
  • Ingestion points: Files (Markdown, HTML) and URLs provided by the user are loaded into the annotation environment (SKILL.md).
  • Boundary markers: The instructions lack delimiters or explicit warnings to disregard potential instructions embedded within the external documents being annotated.
  • Capability inventory: The agent has the capability to execute shell commands and is instructed to "address" annotations returned from the external tool, which could contain malicious instructions (SKILL.md).
  • Sanitization: No validation or sanitization is specified for the input path or URL before it is passed to the shell command execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 9, 2026, 12:02 AM
Security Audit — agent-trust-hub — plannotator-annotate