The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides installation instructions that involve downloading pre-compiled binaries from a personal GitHub repository (JulienTant/blogwatcher-cli) and piping the output to the system's binary directory using curl | tar. This installation method lacks integrity verification and relies on an unverified source, representing a supply chain risk.
[COMMAND_EXECUTION]: The skill's functionality is built around executing shell commands via the blogwatcher-cli tool. The agent is instructed to use these commands to add blogs, scan feeds, and manage articles, which creates an active command execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from RSS/Atom feeds and HTML scraping, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: External content is fetched from arbitrary URLs during feed discovery, scanning, and article listing (e.g., via blogwatcher-cli scan and blogwatcher-cli articles).
Boundary markers: The skill instructions do not specify any delimiters or safety warnings to help the agent distinguish between feed content and operational instructions.
Capability inventory: The skill utilizes the blogwatcher-cli tool which has capabilities for network access and local file system interaction (managing a SQLite database at ~/.blogwatcher-cli/).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external feeds before it is presented to the agent.

blogwatcher