The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the installation of a Python package directly from a GitHub repository ZIP archive (https://github.com/Panniantong/agent-reach/archive/main.zip). Installing code from unverified third-party repositories allows for the execution of arbitrary setup scripts and code during installation.
[COMMAND_EXECUTION]: The command agent-reach install --env=auto is designed to automatically detect the host environment and install multiple system dependencies and CLI tools (Node.js, mcporter, bird CLI, gh CLI). This automated provisioning of executables from an untrusted package represents a significant risk to the host environment.
[EXTERNAL_DOWNLOADS]: The skill references and fetches instructions from a non-trusted repository (https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md) and recommends the installation of third-party browser extensions.
[INDIRECT_PROMPT_INJECTION]: The primary function of the skill is to ingest and process data from external, untrusted sources including Twitter/X, Reddit, YouTube transcripts, and arbitrary web pages.
Ingestion points: The agent-reach read <url> and agent-reach search commands in SKILL.md are ingestion points for external content.
Boundary markers: The instructions do not define boundary markers or delimiters to separate ingested content from agent instructions.
Capability inventory: The skill has access to network operations and can execute various sub-commands of the agent-reach CLI tool.
Sanitization: There is no evidence of sanitization or filtering applied to the content retrieved from external URLs before it is processed by the agent.

agent-reach