The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install an external CLI tool (belt-sh/cli) via npx and references installation scripts hosted at raw.githubusercontent.com/inference-sh/skills. These sources are outside of the predefined trusted repositories.
[COMMAND_EXECUTION]: The skill leverages the belt CLI to execute various tasks, including browser automation (infsh/agent-browser) and search operations. The allowed-tools metadata restricts Bash usage to this specific command.
[DYNAMIC_EXECUTION]: A Python script using matplotlib is embedded within the instructions to generate positioning maps. This code is executed at runtime via the infsh/python-executor application.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill uses tavily/search-assistant, exa/search, and tavily/extract to pull content from competitor websites, review platforms (G2, Capterra), and news articles into the agent's context (SKILL.md).
Boundary markers: No specific delimiters or "ignore instructions" warnings are implemented for the ingested data.
Capability inventory: The agent has access to infsh/python-executor and infsh/agent-browser (documented in SKILL.md).
Sanitization: No explicit sanitization or filtering of the external search results is described.
Note: Processing untrusted external data from the web creates a vulnerability surface where malicious instructions embedded in websites could influence the agent's behavior.

competitor-teardown