Outlook Delegate

Read, search, and manage Outlook emails and calendar via Microsoft Graph API with delegate support. Supports sending as self, as owner (Send As), and on behalf of owner (Send on Behalf). Modified for delegate access from https://clawhub.ai/jotamed/outlook

This skill appears to implement delegate Outlook access as described, but it requires you to create and store sensitive credentials (client_id/client_secret/refresh tokens) in ~/.outlook-mcp and to grant high-privilege Exchange permissions (FullAccess and either SendAs or SendOnBehalf). Before installing: 1) Review the scripts yourself (they call only Microsoft endpoints) and confirm you trust the source. 2) Do not grant both SendAs and SendOnBehalf — pick one per the docs. 3) Use a dedicated, limited 'assistant' account (not an admin or personal account). 4) Limit client secret lifetime, rotate credentials, and monitor mailbox/audit logs. 5) Because the registry metadata omits the required config/credentials, ask the maintainer to update metadata to explicitly declare the config path and secrets required. If you are not comfortable granting Full Mailbox Access or handling OAuth secrets, do not install/use this skill.

Type: OpenClaw Skill Name: outlook-delegate Version: 1.1.1 The OpenClaw AgentSkills skill bundle for Outlook delegate access is classified as benign. The code and documentation consistently align with the stated purpose of managing Outlook emails and calendars via Microsoft Graph API with delegate support. Key security indicators include robust input sanitization using `jq` for JSON payloads and URL encoding for query parameters (e.g., in `scripts/outlook-mail.sh` and `scripts/outlook-calendar.sh`), secure handling of sensitive credentials (client secret via stdin, access token via temporary files with `chmod 600` in `scripts/outlook-token.sh`), and strict file permissions for configuration and credential directories (`~/.outlook-mcp/`). All network communications are directed to legitimate Microsoft OAuth and Graph API endpoints. There is no evidence of data exfiltration, unauthorized remote execution, persistence mechanisms, obfuscation, or prompt injection attempts within the `SKILL.md` or other documentation files. The broad permissions requested are necessary for the skill's stated delegate functionality.