The Agent Skills Directory

Command Execution (MEDIUM): Argument injection vulnerability in scripts/fetch_feeds.py and scripts/fetch_github_issues.py. These scripts call curl via subprocess.run and pass user-controlled or external URL strings directly as arguments without the -- separator. A malicious URL starting with a dash (e.g., -o /tmp/malicious.sh) could be interpreted as a command-line flag by curl, enabling arbitrary file writes or other unintended behavior.
Indirect Prompt Injection (LOW): The skill processes untrusted external data which is then summarized by the agent, creating a risk of indirect prompt injection. 1. Ingestion points: RSS/Atom feed content in fetch_feeds.py and GitHub issue/comment text in fetch_github_issues.py. 2. Boundary markers: Absent. The content is returned as raw text in a JSON structure without instructions for the agent to ignore embedded commands. 3. Capability inventory: The skill possesses network access (curl) and the ability to read local configuration files. 4. Sanitization: Basic HTML tag removal and character truncation are performed, but no natural language instruction filtering is present.
External Downloads (SAFE): The skill's primary function is to fetch remote content from user-specified or hardcoded URLs, which is handled via standard subprocess calls to curl.

feed-digest