seo-images

Warn

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a local script python3 scripts/pagespeed_check.py and refers to bash extensions/firecrawl/install.sh. Since these scripts are not included in the provided skill files, their behavior cannot be audited or verified for safety.
  • [COMMAND_EXECUTION]: The skill logic involves interpolating user-provided URLs into shell commands (e.g., python3 scripts/pagespeed_check.py "{url}"). If the agent fails to sanitize these inputs, it could lead to command injection vulnerabilities.
  • [CREDENTIALS_UNSAFE]: The skill explicitly reads configuration and API keys from ~/.config/seo-skills/google-api.json. Accessing sensitive files within the user's home directory is a high-risk operation.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it scrapes and processes raw HTML and JSON-LD from external websites.
  • Ingestion points: HTML attributes (alt, srcset) and JSON-LD blocks scraped via Firecrawl.
  • Boundary markers: None. There are no instructions to the agent to treat scraped content as untrusted or to ignore embedded instructions.
  • Capability inventory: Subprocess execution (python3) and file writing (generating audit reports).
  • Sanitization: No sanitization of scraped text is documented before it is synthesized into final reports or suggested remediation snippets.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 25, 2026, 01:08 PM
Security Audit — agent-trust-hub — seo-images