The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's setup process (/seo setup) involves downloading and installing an isolated Python runtime and the Chromium browser to the host system from non-standard sources.
[COMMAND_EXECUTION]: The skill uses a custom launcher (claude-seo run) to execute various bundled Python scripts, such as google_auth.py, drift_history.py, and google_report.py, which are part of the skill's internal orchestration logic.
[DATA_EXFILTRATION]: The skill interacts with numerous external service providers, sending target URLs and site metadata to third-party APIs including Moz, Bing, Google (GSC, PageSpeed), DataForSEO, and Geoapify. This transmits site structure and performance data to external entities as part of its auditing function.
[PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection because its primary function is to crawl and analyze untrusted content from external websites. Malicious instructions placed on an audited website could be ingested and followed by the agent during processing.
Ingestion points: Website content is ingested and analyzed through commands like /seo audit and /seo page targeting external URLs.
Boundary markers: The orchestration logic does not describe any use of delimiters or 'ignore embedded instructions' warnings to prevent the agent from obeying instructions found within analyzed HTML content.
Capability inventory: The skill possesses significant capabilities, including shell command execution (claude-seo run), network access to various external APIs, and local file management for baseline drift monitoring.
Sanitization: There is no evidence in the provided files that external content is sanitized or filtered for instructions before being processed by the agent.

seo