rust-daily
Pass
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes content from untrusted external sources (Reddit) without using sanitization or boundary markers.
- Ingestion points: Fetches data from
reddit.com,this-week-in-rust.org,blog.rust-lang.org, andrustfoundation.orgusingagent-browserandWebFetch. - Boundary markers: The instructions lack specific delimiters or instructions to treat the fetched content as data rather than instructions.
- Capability inventory: The skill uses the
agent-browserCLI,mcp__actionbooktools, andWebFetchfor network and browser operations. - Sanitization: No sanitization or validation of the fetched external content is performed before processing.
- [DATA_EXFILTRATION]: The skill accesses the local file system using relative path traversal (
../../agents/rust-daily-reporter.md) to determine its execution mode. While used for configuration checks, accessing files outside the immediate skill directory is a noteworthy behavior. - [COMMAND_EXECUTION]: The skill executes shell commands via the
agent-browserCLI to open URLs and extract text from web pages. - [EXTERNAL_DOWNLOADS]: Fetches news and community updates from well-known sources including Reddit, the official Rust blog, and the Rust Foundation website.
Audit Metadata