The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install external command-line tools and packages from the NPM registry using npx skills add belt-sh/cli and npx skills add inference-sh/skills.
[DYNAMIC_EXECUTION]: The skill generates and executes Python code at runtime using a remote application executor (infsh/python-executor) to create data visualizations (charts) based on user-provided metrics.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection by ingesting data from external search providers like Tavily and Exa.
Ingestion points: Web search results retrieved via tavily/search-assistant, exa/search, and exa/answer as described in SKILL.md.
Boundary markers: The instructions do not define boundary markers or delimiters to help the agent distinguish between its instructions and content returned from the search engines.
Capability inventory: The agent has shell access via the Bash tool and can execute Python scripts via the infsh/python-executor app.
Sanitization: No evidence of input validation, filtering, or sanitization of search results is present in the skill instructions.
[COMMAND_EXECUTION]: The skill makes extensive use of the belt CLI tool to perform network-based search queries, user authentication (belt login), and execution of external logic.

case-study-writing