The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute a local installation script (./extensions/dataforseo/install.sh) and a cost-management Python script (python scripts/dataforseo_costs.py). These scripts are not included in the skill payload, meaning the agent is instructed to run unverified code from the local filesystem.
[COMMAND_EXECUTION]: The cost guardrail logic involves passing dynamic arguments (such as API endpoints and cost values) to the dataforseo_costs.py script. If these arguments are derived from untrusted external data without strict validation, it creates a potential command injection vector.
[EXTERNAL_DOWNLOADS]: The installation process for the DataForSEO extension implies downloading external binaries or scripts from remote locations that are not verified within this skill's instructions.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes large amounts of untrusted data from the live web.
Ingestion points: Untrusted data enters the agent context via SERP results, backlink profiles, on-page content parsing, and the ai_optimization_chat_gpt_scraper tool, which scrapes live ChatGPT search results.
Boundary markers: The instructions do not define clear boundaries (such as delimiters or specific 'ignore instructions' warnings) to prevent the agent from being influenced by malicious content embedded in the fetched search results.
Capability inventory: The skill possesses the capability to execute shell commands and Python scripts, which could be abused if an injection attack succeeds.
Sanitization: There is no evidence of sanitization or filtering being applied to the external data before it is presented to the agent for analysis.

seo-dataforseo