Meeting Transcripts

Capture meeting transcripts from Fireflies.ai via polling or webhooks. Auto-fetches transcripts, extracts action items/decisions/summaries, and writes struct...

This skill appears to do what it says: it reads a Fireflies API key (stored in ~/.openclaw/secrets/fireflies-api-key.txt), polls or accepts webhooks from Fireflies, and writes meeting markdown files to your workspace. Before installing: (1) Be aware the registry metadata does not list the required API key — you must create and store it as the README instructs. (2) The webhook mode requires exposing a local HTTP endpoint (ngrok/Cloudflare Tunnel); only do that if you understand the risks and have protected the endpoint with the optional webhook secret. (3) Store the API key/secret with restrictive file permissions and prefer the webhook secret so the server verifies signatures. (4) Inspect logs/output locations (~/clawd or OPENCLAW_WORKSPACE) for sensitive transcript content and consider access controls for that directory. If you'd like, I can suggest a safer webhook deployment pattern (e.g., host the receiver in a secured cloud function or use a vetted tunneling setup) or generate exact commands to set file permissions and configure the webhook secret.

Type: OpenClaw Skill Name: meeting-transcripts Version: 1.0.0 The OpenClaw AgentSkills bundle for 'meeting-transcripts' is benign. All scripts (`fetch-transcript.js`, `poll-new-meetings.js`, `webhook-server.js`) interact exclusively with the legitimate Fireflies.ai GraphQL API (`https://api.fireflies.ai/graphql`) and store data in the designated `memory/meetings` directory. API keys and webhook secrets are correctly loaded from `~/.openclaw/secrets/`. The `webhook-server.js` script includes HMAC-SHA256 signature verification, a strong security practice. The `SKILL.md` instructions are clear, directly align with the skill's purpose, and contain no evidence of prompt injection attempts or other malicious directives for the AI agent.