The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from the external forum ibbs.pro, which serves as an attack surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent context through results from the ibbs_pro__get_search and ibbs_pro__read_page tools described in SKILL.md.
Boundary markers: The instructions do not define clear delimiters or include warnings to the agent to disregard instructions embedded within the forum content.
Capability inventory: The skill includes powerful tools for browser automation (unbrowse.browse.act), task execution (unbrowse.run), and general scraping, which could be abused if an injection is successful.
Sanitization: No sanitization or validation of the retrieved forum content is mentioned, allowing raw external data to be interpolated into the agent's prompt.
[EXTERNAL_DOWNLOADS]: The skill makes requests to the vendor's infrastructure to fetch configurations and execute tools.
Evidence: The skill fetches OpenAPI specifications and MCP configurations from https://unbrowse.ai/api/... and https://unbrowse.ai/mcp/... as seen in README.md and server.json. These represent standard vendor-to-platform communication.
[COMMAND_EXECUTION]: The documentation provides standard installation commands for the agent environment.
Evidence: README.md provides claude mcp add and npx skills add commands to integrate the skill and its remote MCP server.

ibbs-pro