The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web data using automated search and browsing tools, which creates a potential surface for indirect prompt injection. * Ingestion points: The skill uses 'belt app run' with search engines and browsing tools to ingest content from competitor websites, reviews, and news sources. * Boundary markers: The instructions lack explicit delimiters or instructions to ignore potential commands embedded within the retrieved web data. * Capability inventory: The skill has access to the belt CLI, which includes capabilities for Python execution and shell operations. * Sanitization: There is no documented process for sanitizing or validating external content before it is incorporated into the agent's context or further operations.
[DYNAMIC_EXECUTION]: The skill uses a Python execution tool to run dynamically generated code for creating positioning maps. * Evidence: SKILL.md contains a Python script template utilizing matplotlib which is passed to 'infsh/python-executor' via the belt CLI.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill depends on an external CLI tool to perform its primary functions. * Evidence: The documentation instructs the user to install 'belt-sh/cli' using 'npx skills add'.

competitor-teardown