agent-teams

Interact with Microsoft Teams - send messages, read channels, manage reactions

This skill does what it claims (control Teams), but it obtains credentials by reading the Teams desktop app's local cookie DB and stores a plaintext token in ~/.config/agent-messenger/teams-credentials.json. It also instructs the agent to read/write a persistent memory file at the start of each task. Before installing, consider: 1) Do you trust the npm package author (agent-messenger)? Inspect the package source on the registry or GitHub first. 2) Are you comfortable granting the tool access to app data (and possibly Full Disk Access on macOS)? That allows it to extract session tokens which effectively act as your account credentials. 3) If you proceed, run it from a dedicated account or isolated environment, limit filesystem permissions, and remove the credentials file when done. 4) Prefer OAuth/official integrations where possible; if your organization forbids using user tokens for automation, avoid this skill. Finally, consider asking the author to declare the config paths and credential behavior in the registry metadata so the access surface is explicit.

Type: OpenClaw Skill Name: peng-agent-teams Version: 1.10.6 The skill bundle provides a CLI for Microsoft Teams that authenticates by extracting session tokens (skypetoken_asm) directly from the local Teams desktop application's SQLite cookie database across macOS, Linux, and Windows (SKILL.md, references/authentication.md). While this high-risk credential extraction is the stated purpose for enabling 'seamless' authentication, accessing another application's private session data is a significant security risk. The bundle also includes instructions for the AI agent to manage a local persistent memory file (~/.config/agent-messenger/MEMORY.md), though it contains explicit safety directives to avoid storing sensitive credentials in that file.