The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external URLs which is then processed by the agent using powerful tools.
Ingestion points: Website HTML, sitemaps, and robots.txt files are fetched via WebFetch and curl in the SKILL.md defined orchestration logic.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore potential commands embedded within the retrieved HTML content.
Capability inventory: The skill has access to Bash, WebFetch, and Write tools, which are used to generate synthesized reports and PDF deliverables.
Sanitization: There is no mention of sanitizing external content before it is interpolated into analysis prompts or report generation logic.
[REMOTE_CODE_EXECUTION]: The /geo update command is designed to pull the latest skill updates from an "upstream" repository. The absence of explicit source verification or integrity checking in the configuration poses a risk of remote code execution if the update source is compromised.
[COMMAND_EXECUTION]: The skill leverages the Bash tool to perform complex system operations, including a multi-step pipeline that uses pandoc and a headless Google Chrome instance to convert markdown reports into PDFs.
[EXTERNAL_DOWNLOADS]: The skill performs automated network requests to fetch content from arbitrary, user-provided URLs for auditing and citability scoring.

geo