The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading the blogwatcher-cli tool via go install, docker pull from GitHub Container Registry (GHCR), and direct binary downloads from GitHub releases.
Evidence: curl -sL https://github.com/JulienTant/blogwatcher-cli/releases/latest/download/... and ghcr.io/julientant/blogwatcher-cli.
These downloads are from well-known platforms (GitHub) and target the official repository of the tool described, representing standard installation practices.
[COMMAND_EXECUTION]: The skill documents numerous CLI commands for the blogwatcher-cli tool, including commands to add blogs, scan feeds, and manage article read states.
Evidence: blogwatcher-cli add, blogwatcher-cli scan, blogwatcher-cli articles, etc.
These commands are consistent with the skill's stated purpose of monitoring blog feeds.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from RSS/Atom feeds and HTML scraping, which could theoretically contain malicious instructions.
Ingestion points: Reads data from external URLs via RSS/Atom feeds and HTML scraping (SKILL.md).
Boundary markers: None explicitly documented in the instructions.
Capability inventory: Uses a CLI tool (blogwatcher-cli) to interact with a local SQLite database and external web services; the agent itself is intended to run these commands.
Sanitization: Not specified in the skill documentation, relying on the underlying blogwatcher-cli tool's handling of feed data.
As with any feed reader, there is a surface for indirect injection if the agent is asked to process or summarize the content of an untrusted feed.

blogwatcher