The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze data from multiple external sources, which constitutes an attack surface where malicious instructions could be embedded in the processed data.
Ingestion points: The agent is instructed to fetch and process data from external services including DataForSEO, SerpApi, Google Search Console, and various AI search engines (ChatGPT, Perplexity, Claude) as specified in references/full-sop.md.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" guards to isolate untrusted external content from the agent's core instructions.
Capability inventory: The skill possesses capabilities that could be exploited if an injection occurs, such as executing shell commands (curl) for the IndexNow API and potentially performing automated deployments to platforms like GitHub or Vercel (referenced in Section 12.3 of references/full-sop.md).
Sanitization: There is no mention of sanitizing, escaping, or validating the data retrieved from external APIs or web pages before it is processed by the agent.
[COMMAND_EXECUTION]: The skill includes a pre-formatted curl command template in section 9 of references/full-sop.md used for interacting with the IndexNow API. While this targets a well-known service (indexnow.org) and is a legitimate SEO practice, it represents a command execution capability.
[SAFE]: The skill follows security best practices regarding secret management by explicitly instructing users to store API keys and tokens in environment variables or secure vaults rather than hardcoding them in the skill files.

gr-seo-geo-agent