securityclaw-autonomous-soc-agent
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill includes instructions to download and install the Ollama service using a shell script fetched from
ollama.com. - [DYNAMIC_EXECUTION]: The agent architecture relies on dynamically discovering and executing Python code from files within the
skills/directory to perform specific security functions. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted security logs and external threat intelligence, which presents a potential vulnerability to indirect prompt injection.
- Ingestion points: Untrusted data enters the agent context via OpenSearch database queries and external API responses (e.g., AbuseIPDB) as documented in the
logic.pyexamples inSKILL.md. - Boundary markers: Example prompt templates do not demonstrate the use of delimiters or specific instructions for the agent to ignore commands embedded within log data.
- Capability inventory: The agent possesses capabilities to execute database searches, perform network operations to external APIs, and maintain persistent conversation history in a SQLite database.
- Sanitization: Logic examples show direct interpolation of event data into LLM prompts using f-strings without explicit sanitization or filtering mechanisms.
Recommendations
- HIGH: Downloads and executes remote code from: https://ollama.com/install.sh - DO NOT USE without thorough review
Audit Metadata