The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides diagnostic tools that fetch external HTML content via curl and extract JSON-LD blocks for analysis. This introduces a vulnerability surface where a malicious website could embed prompt injection payloads within its structured data to influence the agent's reasoning or subsequent actions.
Ingestion points: Shell commands in SKILL.md used for extraction (e.g., curl -s ... | python3 ...).
Boundary markers: Absent. The agent is directed to read and analyze the raw JSON output from the fetched URL.
Capability inventory: Subprocess execution for shell commands including curl and python3.
Sanitization: Absent. The provided Python script parses and prints data directly without sanitizing it for downstream LLM consumption.
[DYNAMIC_EXECUTION]: The skill utilizes python3 -c to execute inline Python scripts provided within the documentation. While the specific scripts provided for SEO auditing are currently benign, the use of dynamic execution placeholders is a potential vector for executing unverified code if the context is modified.
[COMMAND_EXECUTION]: The diagnostic section includes complex shell one-liners involving pipes, environment variables (user-agents), and redirection. These commands are intended for technical auditing but require the agent to operate with shell execution privileges to perform network requests and data processing.

schema-markup