The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the automated download of a large collection of playbooks and operational logic from a third-party GitHub repository (https://github.com/chadboyda/agent-gtm-skills). The source organization is not categorized as a trusted vendor.
[REMOTE_CODE_EXECUTION]: The installation instructions utilize npx skills add to dynamically fetch and integrate remote content into the agent's execution environment. This mechanism executes code retrieved from the external repository during the installation phase.
[COMMAND_EXECUTION]: The skill body contains shell scripts designed to check for the existence of files and grep through local skill definitions to determine installation status. It also explicitly encourages the use of ! syntax for direct shell command execution.
[INDIRECT_PROMPT_INJECTION]: The skill represents a significant attack surface for indirect injection by processing external marketing data, lead enrichment info, and GTM metrics.
Ingestion points: Processes user-provided product positioning data, pricing strategies, and lead lists across 18 different sub-skills.
Boundary markers: No specific delimiters or safety instructions are present in the provided stubs to isolate untrusted data from the agent's instructions.
Capability inventory: The skill possesses the ability to execute shell commands (bash), install packages (npx), and perform file system checks.
Sanitization: No evidence of input validation or sanitization of ingested data before interpolation into agent prompts.

chadboyda-gtm