proofread

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation in SKILL.md provides specific shell commands for the agent to execute when performing a Council proofread. It directs the agent to run uv run python -m cli_council using various file paths. While this is a documented feature for multi-model synthesis using the vendor's own packages, the execution of shell commands increases the risk of command manipulation if the agent is influenced by malicious input.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes user-provided LaTeX source code and log files. * Ingestion points: The skill uses Read, Glob, and Grep tools to access all .tex and .log files in a project directory. * Boundary markers: No delimiters or ignore instructions are specified to separate untrusted document content from the agent's analytical instructions. * Capability inventory: The skill has permissions for file system access and the ability to execute shell commands via the council CLI. * Sanitization: There is no evidence of content sanitization or validation before the LaTeX data is included in prompts for the agent or passed to the external CLI tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 07:18 PM