The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/slack_client.py utilizes subprocess.Popen in the trigger_background_user_refresh function. It executes a new instance of itself to refresh the user cache in the background. While the arguments are passed as a list and shell=True is not used, this constitutes dynamic process execution.
[PROMPT_INJECTION]: The skill processes untrusted external data in the form of Slack messages and search results. These are summarized and presented to the agent, creating a surface for indirect prompt injection. A malicious message could contain instructions designed to influence the agent's behavior.
Ingestion points: Data enters via conversations_history, conversations_replies, and search_messages functions in scripts/slack_client.py.
Boundary markers: The instructions in SKILL.md do not specify the use of delimiters or system instructions to ignore potential commands embedded in Slack messages.
Capability inventory: The skill has the ability to send messages to Slack, search history, and export data to the local file system.
Sanitization: No explicit sanitization or filtering of Slack message content is implemented in the provided Python script.

slack