Skill v1.0.0
currentTrusted Publisher100/100version: "1.0.0" user-invocable: false description: "DEPRECATED: Use /copilot-studio:chat-with-agent instead — it auto-detects DirectLine vs M365 mode. This skill is kept for backwards compatibility only." argument-hint: <utterance to send> allowed-tools: Bash(node directline-chat.bundle.js ), Bash(node manage-agent.bundle.js detect-mode ), Read, Glob, Grep context: fork agent: copilot-studio-test
DirectLine Chat (Deprecated)
This skill has been merged into `/copilot-studio:chat-with-agent`, which now auto-detects the agent's authentication mode and uses DirectLine automatically for no-auth and manual-auth agents.
Use `/copilot-studio:chat-with-agent` instead. It handles both DirectLine and Copilot Studio SDK modes in a single skill.
If you are here because the caller explicitly provided a DirectLine secret or token endpoint URL, you may proceed with the instructions below. Otherwise, redirect to /copilot-studio:chat-with-agent.
Direct Usage (when caller provides explicit DirectLine credentials)
Token endpoint mode
node ${CLAUDE_SKILL_DIR}/../../scripts/directline-chat.bundle.js \--token-endpoint "<url>" "<utterance>"
DirectLine secret mode
node ${CLAUDE_SKILL_DIR}/../../scripts/directline-chat.bundle.js \--directline-secret "<secret>" "<utterance>"
Multi-turn
node ${CLAUDE_SKILL_DIR}/../../scripts/directline-chat.bundle.js \--token-endpoint "<url>" "<follow-up>" \--conversation-id <id> --directline-token "<token>"
Always pass `--directline-token` when resuming. DirectLine tokens are bound to the conversation. They expire after ~30 min.
Output format
Same as /copilot-studio:chat-with-agent Phase 1a — see that skill for full documentation of JSON output, sign-in flow, and error handling.