The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The references/install.md file provides instructions for the user to clone the skill's repository from GitHub (https://github.com/JasonColapietro/suede-creator-skills.git) and execute an install.sh script. These instructions facilitate the fetching of code from an external repository provided by the vendor.
[REMOTE_CODE_EXECUTION]: The documentation describes patterns for downloading and executing remote scripts using bash and a Python-based installer (install-skill-from-github.py). These patterns allow for the execution of code hosted on the vendor's GitHub repository within the host environment.
[COMMAND_EXECUTION]: The skill documentation references the execution of shell commands for installation and setup, which involves direct interaction with the system's command-line interface.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to ingest and analyze data from untrusted external sources.
Ingestion points: The skill's workflows (e.g., Suedify, SEO Audit, and Code Review) involve reading content from live URLs and inspecting code repository files as described in SKILL.md and references/condensed-workflows.md.
Boundary markers: There are no explicit instructions or delimiters defined to isolate content retrieved from external sources from the agent's operational instructions.
Capability inventory: The umbrella skill routes requests to specialized agents with the ability to perform multi-file repository modifications and code analysis.
Sanitization: The instructions do not define methods for sanitizing or filtering external content before it is processed by the agent.

suede-workflow-skills