The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large volumes of unstructured data from Slack, Google Drive, Email, and Calendar events to summarize internal updates, newsletters, and FAQs. This creates a significant surface for indirect prompt injection attacks.
Ingestion points: Data is fetched from Slack messages, Google Drive documents, emails, and calendar attachments as described in examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between the system's instructions and potentially malicious instructions embedded within the ingested data.
Capability inventory: The skill uses the agent's core capabilities to read external data and format it into user-facing communications. While it does not explicitly call shell commands or write files in the provided scripts, the formatted output could be used to influence subsequent agent actions in a tool-chaining scenario.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content pulled from these external sources before it is processed by the LLM.

internal-comms