The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/setup.sh script downloads a pre-built CLI binary from the official vendor repository (github.com/getdex/agent-skills). It also retrieves the clihub utility using go install and the add-mcp tool via npx. These resources are used to establish the connection between the AI agent and the Dex service.\n- [REMOTE_CODE_EXECUTION]: The setup process involves downloading and executing remote code, specifically a pre-compiled binary from GitHub and external packages from the Go and NPM registries. This is a standard part of the skill's installation and authentication process.\n- [COMMAND_EXECUTION]: The setup.sh script executes shell commands to perform environment detection (OS, architecture, SSH status) and runs the dex CLI to facilitate user authentication.\n- [CREDENTIALS_UNSAFE]: The skill manages sensitive API keys and OAuth tokens, which are stored in local configuration files such as ~/.dex/api-key and ~/.clihub/credentials.json. It also utilizes the DEX_API_KEY environment variable for headless environments.\n- [PROMPT_INJECTION]: The skill is subject to an indirect prompt injection surface as it processes untrusted contact data and interaction notes from the CRM.\n
Ingestion points: Data retrieved from contact records and timeline notes via dex_search_contacts and dex_get_contact.\n
Boundary markers: The skill does not currently use specific delimiters or instructions to ignore embedded commands in the contact data.\n
Capability inventory: The skill has the capability to delete contacts, merge records, and update timeline notes, which could be targeted by malicious content within the CRM.\n
Sanitization: No explicit sanitization or filtering of the ingested contact data is performed in the provided skill scripts.

dex-skill