mxs-cli-ai-author

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill's documentation specifies the default file system locations for sensitive configuration and authentication data, specifically ~/.config/mxs/credentials.json (containing access tokens and refresh tokens) and ~/.config/mxs/config.json (documented in references/auth-config-output.md). These are vendor-specific paths managed by the mxs tool.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of reading and writing content from an external CMS.
  • Ingestion points: The agent is instructed to fetch potentially untrusted data from the server using commands like mxs post get, mxs note get, and mxs comment list (found in references/command-index.md).
  • Boundary markers: The instructions do not define strict delimiters or boundary markers to distinguish content retrieved from the server from the agent's internal instruction sequence.
  • Capability inventory: The agent has a broad set of capabilities, including the ability to delete content (mxs post delete) and modify server-wide settings (mxs config set), which could be targeted by instructions embedded in the ingested data (found in references/command-index.md).
  • Sanitization: No specific requirements are provided for sanitizing or escaping the content read from the server before the agent processes it.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 08:43 PM
Security Audit — agent-trust-hub — mxs-cli-ai-author