cloud-saas-exposure
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external sources, creating a potential surface for indirect prompt injection.\n
- Ingestion points: The skill parses package manifest files (e.g., package.json, requirements.txt, .npmrc) and structured XML/JSON responses from cloud storage bucket listings.\n
- Boundary markers: There are no explicit instructions or delimiters defined to isolate the content of these external files from the agent's control flow.\n
- Capability inventory: The skill utilizes shell commands (curl, grep) and a Python interpreter to process data and make network requests.\n
- Sanitization: While the skill uses specific patterns (regex) for data extraction, it does not explicitly define sanitization or validation logic to prevent malicious payloads within these files from influencing the agent's behavior.\n- [DYNAMIC_EXECUTION]: The skill provides a standalone Python script (§7.4) designed to be executed by the agent to perform offline decoding of AWS account identifiers. This script uses only standard libraries and performs deterministic bitwise operations, but represents a pattern of providing source code for runtime execution.\n- [COMMAND_EXECUTION]: The skill includes shell command recipes for performing network probes and candidate generation. These commands are localized to the agent's environment and target well-known cloud provider endpoints and package registries.
Audit Metadata