The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources that are not under the direct control of the user or the agent. This creates a potential surface for indirect prompt injection where malicious instructions could be embedded in external content.
Ingestion points: The skill ingests untrusted data from the web via tools such as get_serp_results, get_google_business_questions, and get_ranked_keywords (which pulls data from target domains).
Boundary markers: The instructions do not explicitly define boundary markers (like XML tags or delimiters) to wrap the external data, though it advises presenting data in structured tables.
Capability inventory: The skill has the capability to modify state using update_project_context, save_keywords, and appendResearchLog.
Sanitization: There is no explicit mention of sanitization or filtering of the content retrieved from external tools before it is processed or presented to the user.
[SAFE]: The skill demonstrates safe operational practices by requiring explicit user confirmation before executing persistent actions like save_keywords. It also includes an optimization check (the 30-day research log check) to prevent unnecessary credit consumption, which is a benign and helpful feature for the user.

keyword-research