The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize data from platforms that can be controlled by third parties (Slack, Email, Google Drive, Calendar).
Ingestion points: examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md instruct the agent to retrieve updates and documents from these collaborative tools.
Boundary markers: The guidelines do not provide delimiters or "ignore previous instructions" warnings to prevent the agent from obeying commands hidden in the ingested text.
Capability inventory: The skill possesses the capability to aggregate sensitive internal information into summaries and newsletters, which could be manipulated by an attacker who can post to Slack or edit shared documents.
Sanitization: No sanitization or validation logic is included to ensure that content from external platforms does not contain malicious prompt fragments.

internal-comms