skills/shhac/agent-skills/agent-slack/Gen Agent Trust Hub

agent-slack

Warn

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATIONCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill includes functionality to extract Slack authentication tokens (xoxc and xoxd) from sensitive local file paths, including Chrome, Brave, Firefox, Zen, Opera, and Safari browser profiles and cookie stores (documented in references/commands/auth.md). While this is intended for authentication, the ability to harvest browser data is a significant capability.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from Slack messages, threads, and search results without explicit boundary markers or sanitization, creating a surface where instructions from external Slack users could influence the agent's behavior. \n
  • Ingestion points: Commands such as message get, message list, search messages, and unreads (references/commands/message.md, references/commands/search.md) bring external content into the agent's context. \n
  • Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in fetched content. \n
  • Capability inventory: The skill provides extensive write capabilities including message send, edit, delete, react, channel invite, and workflow run (references/commands/message.md, references/commands/channel.md). \n
  • Sanitization: There is no documentation regarding the sanitization or filtering of external content before processing.- [PRIVILEGE_ESCALATION]: On macOS, the skill's auth import-browser safari command requires the user to grant "Full Disk Access" to enable the extraction of tokens from the Safari cookie store (references/commands/auth.md).- [COMMAND_EXECUTION]: The api call command serves as a raw escape hatch to the Slack Web API, allowing the agent to execute any API method. This bypasses the higher-level command wrappers and their associated safety gates, such as the --yes confirmation requirement (references/commands/other.md).- [CREDENTIALS_UNSAFE]: The skill manages sensitive Slack authentication secrets in local storage (credentials.json) and supports the use of environment variables like SLACK_TOKEN and SLACK_COOKIE_D for session persistence (references/commands/auth.md, references/commands/cache-config.md).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 31, 2026, 09:40 PM
Security Audit — agent-trust-hub — agent-slack