client-secret-exposure-audit
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from remote, user-supplied URLs, creating a surface for indirect prompt injection attacks.
- Ingestion points: The skill fetches HTML and JavaScript bundles from a target URL using curl and stores them in temporary files (e.g., body.html, js_*) for analysis.
- Boundary markers: There are no delimiters or explicit instructions to the agent to treat the fetched content as untrusted data, which could allow malicious content to influence agent behavior.
- Capability inventory: The agent utilizes shell commands (curl, grep, sed, tr) and local file system access to perform reconnaissance.
- Sanitization: The skill lacks mechanisms to sanitize or escape the fetched remote content before it is triaged and scored by the agent.
- [COMMAND_EXECUTION]: The skill's primary audit workflow is implemented through the execution of shell commands, including network operations and complex text processing with grep and sed. While these commands are necessary for the skill's stated purpose, they represent a capability for the agent to execute shell commands based on external data.
Audit Metadata