The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external web sources via search engines and web extraction tools. This creates an attack surface where instructions hidden within third-party websites or search results could potentially influence the agent's behavior.
Ingestion points: Data returned from tavily/search-assistant, exa/search, tavily/extract, and infsh/agent-browser calls within SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard embedded commands in the ingested data.
Capability inventory: The skill utilizes shell command execution (belt CLI), remote Python execution, and network-based search tools.
Sanitization: No evidence of data sanitization or validation is present in the skill's instructions.
[EXTERNAL_DOWNLOADS]: The skill references and points to installation instructions hosted on a GitHub repository (inference-sh/skills). This is a standard practice for documenting dependencies on well-known version control platforms.
[DYNAMIC_EXECUTION]: The skill employs a remote Python execution environment (infsh/python-executor) to generate positioning maps. While the example provided uses standard visualization libraries, this interface allows for the assembly and execution of scripts at runtime.

competitor-teardown