geo-content-optimizer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions specify the execution of a local Python script to perform citation analysis.\n
- Evidence: In
SKILL.md, the skill describes a 'Keyless AI-citation probe' using the commandpython3 \"${CLAUDE_PLUGIN_ROOT}/scripts/connectors/tavily.py\" search \"<target query>\" --answer --limit 10.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external text and URLs provided by users, creating a surface for indirect prompt injection attacks where malicious instructions in the source content could manipulate the agent's behavior.\n - Ingestion points: Content URLs, raw text, and competitor examples are ingested as primary inputs in
SKILL.mdandreferences/ai-overview-recovery.md.\n - Boundary markers: No delimiters or specific 'ignore' instructions are defined for processing external data.\n
- Capability inventory: The skill has the capability to execute shell commands (via the Tavily probe) and write persistent data to the agent's memory directories (
memory/content/,memory/hot-cache.md).\n - Sanitization: There is no evidence of input validation or sanitization for the content processed during the optimization steps.
Audit Metadata