The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's installation process involves downloading code from an untrusted GitHub repository and executing it within the user's environment.
Evidence: Instructions to git clone from https://github.com/MagicStarfishBoost/r15-shanraisshan-claude-code-best-practice-seo.git and then register the skill using /read.
Evidence: Option to download a ZIP archive directly via curl from the same untrusted source.
Evidence: The troubleshooting section suggests running pip install -r requirements.txt and python scripts/authenticate_gsc.py, which would execute any code contained in the downloaded repository.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from arbitrary external URLs, creating a significant attack surface for indirect prompt injection.
Ingestion points: Multiple commands such as /content-audit, /technical-seo, /competitor-gap, and /page-speed-seo fetch content from user-specified websites.
Boundary markers: The skill does not define any delimiters or instructions to prevent the AI agent from following instructions potentially embedded in the processed HTML or content of the audited sites.
Capability inventory: The agent has access to sensitive API keys (Ahrefs, Semrush, etc.) and filesystem paths for Google Search Console credentials. It can also perform network operations and write files to the local system.
Sanitization: There is no mention of sanitizing or validating the data retrieved from external websites before processing it.
[COMMAND_EXECUTION]: The skill encourages the execution of several shell commands that could be exploited if the source repository were compromised.
Evidence: Commands like unzip, cp -r, and pip install are used in the setup and troubleshooting instructions.

seo-content-marketing-claude-skill-suite