The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch content from external URLs and pass the resulting data directly into LLM prompts for analysis and structured data extraction.
Ingestion points: Untrusted content is ingested via curl in the "Page Analysis", "Security Scan", and "Extract Structured Data" sections of SKILL.md.
Boundary markers: The skill uses simple text headers (e.g., "Analyze this webpage:", "HEADERS:", "CONTENT SAMPLE:") but lacks robust delimiters or explicit instructions to the AI to ignore embedded commands within the fetched data.
Capability inventory: The environment has access to curl, bash, python3, and the gemini CLI tool.
Sanitization: No sanitization or filtering is performed on the ingested content before it is interpolated into the prompts.
[DYNAMIC_EXECUTION]: The skill demonstrates the use of python3 -c to execute a multi-line Python script provided as a string literal for HTML parsing.
Evidence: The "Extract Text Content" section pipes curl output to a static Python script defined within the bash command.
Context: This is a low-risk pattern used for data processing rather than executing untrusted code, as the Python logic is hardcoded in the skill and only the input data is dynamic.

spider