The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data which could lead to indirect prompt injection.
Ingestion points: YouTube comments fetched via youtube-video, web page text analyzed by nlp <url>, and data returned from Google Search Console and Indexing APIs.
Boundary markers: The provided documentation does not specify the use of delimiters or instructions to the agent to disregard instructions found within processed data.
Capability inventory: The skill utilizes subprocess execution for its scripts, performs network operations to external Google APIs, and can write report files to the local disk.
Sanitization: Includes a validate_url function to protect against SSRF by blocking loopback and private IP ranges, but does not detail sanitization of retrieved content strings.
[COMMAND_EXECUTION]: The skill operates by executing local Python scripts for nearly all of its functions.
Many commands in SKILL.md rely on executing scripts located in the scripts/ directory, such as python scripts/gsc_query.py and python scripts/pagespeed_check.py.
The integration guide in references/cross-skill-integration.md demonstrates the use of python3 -c to run inline code, specifically for URL validation.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python libraries to function.
Evidence: Documentation in references/keyword-planner-api.md instructs the user to run pip install google-ads.

seo-google