The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill fetches data from attacker-controllable sources such as Reddit post titles and community blogs.
Ingestion points: The skill uses agent-browser and WebFetch to scrape content from reddit.com, this-week-in-rust.org, and rustfoundation.org.
Boundary markers: No explicit delimiters or sanitization logic is present to isolate external content from the agent's instructions.
Capability inventory: The skill possesses the capability to spawn a general-purpose subagent (Task) and execute browser-based CLI commands.
Sanitization: There is no evidence of filtering or escaping logic for the retrieved data before it is formatted and presented to the agent or user.
Path Traversal / Local File Access (MEDIUM): The skill attempts to read execution logic from ../../agents/rust-daily-reporter.md. Accessing files using relative paths outside the immediate skill directory is a risk factor that could be exploited to load unauthorized agent configurations if the environment is not properly sandboxed.
Command Execution (LOW): The skill utilizes the agent-browser CLI and mcp__actionbook tools to perform its tasks. While the target URLs are currently limited to well-known domains, the reliance on shell-like command execution for data retrieval increases the attack surface.

rust-daily