The Agent Skills Directory

[DATA_EXFILTRATION]: The skill documentation in references/slack-tasks.md contains a hardcoded Slack archive URL (https://vercel.slack.com/archives/...). This reference points to a well-known service (Slack) and a trusted organization (Vercel), making it a benign example of workspace interaction documentation.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest and process untrusted text content from Slack messages and threads. This is an inherent risk of the skill's primary function of reading team communications, where malicious instructions could theoretically be embedded in Slack messages.
[COMMAND_EXECUTION]: The skill uses agent-browser through shell commands to perform browser automation tasks such as taking snapshots, clicking elements, and filling forms. These commands are fixed in the instructions and restricted to the agent-browser toolset defined in the frontmatter.

slack