The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large amounts of untrusted data from 14+ external platforms, creating a significant attack surface for indirect prompt injection.
Ingestion points: Data is pulled from Twitter/X, Reddit, YouTube transcripts, GitHub issues, RSS feeds, and general web pages via Jina Reader and Exa search.
Boundary markers: The instructions do not define clear boundaries or delimiters to help the agent distinguish between its system instructions and the data retrieved from external platforms.
Capability inventory: The skill has access to shell execution (curl, yt-dlp, mcporter, xreach), file system persistence (~/.agent-reach/), and Python script execution.
Sanitization: There is no mention of sanitizing or validating the content retrieved from external sources before the agent processes or acts upon it.
[EXTERNAL_DOWNLOADS]: The skill fetches setup and installation instructions from a remote GitHub repository to configure platform channels.
Evidence: Fetches install.md from https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md and instructs the agent that "Everything else is your job" when configuring channels based on this guide.
[COMMAND_EXECUTION]: The skill relies on several CLI tools and shell commands to function.
Evidence: Frequent use of mcporter, xreach, yt-dlp, gh, and curl for interacting with web services and local toolsets.
[DYNAMIC_EXECUTION]: Instructions include Python snippets executed via the command line to process WeChat articles and RSS feeds.
Evidence: Python execution blocks using miku_ai and feedparser libraries are used to handle data processing tasks.

agent-reach