The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill requires a custom setup process via the command /seo setup which downloads and installs an isolated Python runtime and Chromium. This mechanism is used to execute bundled scripts such as google_auth.py, render_page.py, and backlinks_auth.py through a custom launcher claude-seo run.
[EXTERNAL_DOWNLOADS]: The skill downloads binary executables and runtimes during the setup phase. It also interfaces with numerous external APIs including Moz, Bing, Google Search Console, and DataForSEO, and fetches raw content from arbitrary URLs provided by the user.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands for dependency management and analysis, including npx lighthouse and the custom claude-seo binary.
[DYNAMIC_EXECUTION]: The skill utilizes a dynamic execution pattern where scripts are loaded and executed at runtime based on the sub-skill being invoked. This includes the execution of Python scripts from a bundled repository location.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as its primary function is to ingest and process untrusted data from the web.
Ingestion points: Website HTML content, sitemaps, and search engine results fetched via fetch_page.py, render_page.py, and the Firecrawl extension.
Boundary markers: None detected in the instructions for separating analyzed content from agent instructions.
Capability inventory: The agent has the capability to write files (google_report.py), perform network operations (multiple APIs), and execute local commands (claude-seo run).
Sanitization: No explicit sanitization or filtering of fetched web content is described in the provided analysis logic.

seo