The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large volumes of untrusted or externally-sourced data from corporate communication tools (Slack, Email, Google Drive, etc.) to summarize updates or answer FAQs.
Ingestion points: examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md all explicitly instruct the agent to read content from Slack channels, email threads, and shared documents.
Boundary markers: The instructions do not specify any boundary markers (like XML tags or delimiters) or instructions to ignore embedded prompts within the retrieved data.
Capability inventory: The skill primarily performs text processing and summarization. It does not explicitly list tools for writing to files or making network requests, but the allowed-tools field is absent in the frontmatter, meaning the environment may provide these capabilities.
Sanitization: No sanitization or validation of the retrieved text is mentioned before it is processed by the LLM.
[DATA_EXFILTRATION]: The skill encourages the agent to search through sensitive communication channels (~Slack, ~Email, ~Google Drive) for "critical team members," "executive announcements," and "fundraising" information. While intended for internal reporting, there is a risk that this information could be exposed if the agent is prompted maliciously or if the resulting summary is sent to unauthorized parties.

internal-comms