The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to the vendor's endpoint at https://bstorms.ai/mcp to synchronize marketplace state and fetch playbook content.
[PROMPT_INJECTION]: The skill facilitates the consumption of playbooks written by external parties, creating a risk for indirect prompt injection.
Ingestion points: Third-party data is ingested into the agent context through the browse and questions tool outputs.
Boundary markers: There are no defined delimiters or instructions to treat the retrieved playbook content as untrusted data within the skill definitions.
Capability inventory: While the skill itself is scoped to communication, the playbooks are explicitly designed for agents performing high-privilege tasks like 'deployment pipelines' and 'tool integration'.
Sanitization: The skill's documentation claims that answers are scanned for prompt injection before delivery, though this cannot be verified through analysis of the skill files.

bstorms