The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of a third-party CLI tool called treg. It instructs the agent to visit an external URL (https://treg.to/llms.txt) and follow instructions to install and sign into the tool. This represents a dependency on unverifiable external software.
[COMMAND_EXECUTION]: The skill's primary operations involve executing shell commands using the treg toolset. Examples include treg catalog search, treg catalog get, and treg org agent-new. This creates a dependency on the integrity and security of the external treg binary.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from multiple untrusted external platforms, creating a significant attack surface for indirect prompt injection.
Ingestion points: The skill reads and processes posts, comments, reactions, and reviews from LinkedIn, Reddit, X (Twitter), and Google Reviews.
Boundary markers: Absent. There are no instructions to use delimiters or ignore embedded instructions within the fetched social media or review content.
Capability inventory: The skill possesses network capabilities (via treg API calls) and file system access (writing and reading signals.csv).
Sanitization: No sanitization or validation logic is specified for the content retrieved from external social media or review platforms.

lead-signals