agent-slack
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to download an installation script from 'https://raw.githubusercontent.com/stablyai/agent-slack/main/install.sh'. Per the trust guidelines, 'stablyai' is a recognized vendor resource, and this download is documented as a standard setup procedure.
- [REMOTE_CODE_EXECUTION]: The installation command
curl -fsSL https://raw.githubusercontent.com/stablyai/agent-slack/main/install.sh | shexecutes a remote script. While this is a high-risk pattern generally, it is applied here specifically for official vendor software installation and is therefore rated as safe in this context. - [COMMAND_EXECUTION]: The skill uses a CLI tool named
agent-slackto perform various Slack operations (reading history, searching, sending messages). These are standard functional capabilities for a Slack integration. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Slack (message history, search results, thread content). While this creates a surface for indirect prompt injection, the skill includes several mitigation strategies: it recommends using 'safe mode' (
AGENT_SLACK_SAFE_MODE=1) to prevent automated posting, suggests using 'drafts' for human review before sending, and advises the agent to verify resulting JSON metadata for mutations. These are considered positive security practices.
Audit Metadata