robustness-table

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Potential shell command injection in Step 6. The skill instructs the agent to run uv run jupytext --sync notebooks/<name>.md, where <name> is derived from the <notebook> argument. If the input is not sanitized (e.g., test.md; rm -rf /), it could lead to arbitrary command execution.
  • [PROMPT_INJECTION]: The skill reads and interprets content from external notebook files to identify regression specifications. This represents an indirect prompt injection surface where a malicious notebook could attempt to influence the agent's code generation or file writing behavior.
  • Ingestion points: SKILL.md Step 1 (Read specified notebook file).
  • Boundary markers: Absent. The agent is not instructed to treat the notebook content as untrusted data.
  • Capability inventory: Write, Edit, Bash (SKILL.md Steps 3, 4, 6).
  • Sanitization: Absent. There is no mention of validating or escaping the content found within the notebook before using it to generate new code or execute shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 10:10 AM
Security Audit — agent-trust-hub — robustness-table