agent-lab

Generate multi-lens narratives for Bride of Charlie from structured data (episodes, claims, graph, social threads); support claim-constraint relaxation for alternate tellings; align book-app chat with meta-layer sources. Use when authoring BOC stories, book-app copy, or enabling others to narrate from the same evidence base.

Plans and operates Discord for Pacha (family-safe, slate/community) and Meta-Layer (governance, DPs, gov hub handoffs): channel maps, mod norms, escalation, announcements, and what belongs in Discord vs GitHub. Use when the user mentions Discord, server setup, moderation, Pacha Discord, meta-layer Discord, community rules, or bot/workflow hooks for servers.

Drafts Desirable Property text and inscription-ready variants from app.themetalayer.org DP sources, applying the full meta-layer lens. Saves outputs to the agent-lab Draft Editor via save_draft when available. Produces article-first or video-first companion assets and CTAs for comments and supporting/opposing artifacts. Use when working on DP drafts, on-chain inscription text, gov hub copy, meta-layer properties, Draft Editor, or "draft DPn / desirable property / inscription."

Ingest documents from Google Drive into the meta-layer knowledge graph. Converts Word docs, PDFs, and Google Docs to structured knowledge, extracts concepts, and adds to Neo4j. Use when the user wants to process Drive documents, mentions "ingest from Drive", "add Drive docs to graph", or needs to organize Google Drive content into their knowledge base.

Organize Google Drive documents using meta-layer concepts and knowledge graph relationships. Auto-categorizes documents, creates folder structures, generates knowledge maps, and maintains semantic organization. Use when user wants to "organize Drive by concepts", "restructure Drive folders", "categorize documents", or "create knowledge-based Drive organization".

Monitor Google Drive folders for new or updated documents and automatically trigger ingestion workflows. Sets up webhooks and polling to detect changes. Use when user wants to "watch Drive for changes", "auto-ingest new documents", "monitor Drive folder", or set up automated document processing.

Drafts Moltbook (agent-native social) presence for the Meta-Layer: profile framing, first posts, Submolt ideas, and CTAs that hand off to the gov hub / desirable-property drafts and artifact comments. Use when the user mentions Moltbook, Submolts, inviting PAIs, agent communities, or agent-only social outreach.

Runs competitive scoring ("jury") across Pacha film producer packets and slate timeline / canon coherence in projects/pacha. Use when comparing, scoring, picking, or ranking Pacha films or producer drafts, or aligning the slate to Pacha's timeline. For partners, outreach, community ops, content briefs, or community-input synthesis, use pacha-producer-outreach instead.

Runs one bounded "producer shift": evaluate backlog/context, pick the next Pacha slate milestone per the autonomy charter, route to jury or outreach work, produce drafts and a human handoff (PR/issue checklist). Use when the user wants agent-initiated producer work, a daily/weekly slate ritual, autonomous next step, or orchestration across pacha-producer-jury and pacha-producer-outreach without merging or publishing externally.

Partner identification and outreach, social community planning, content briefs, and community-input synthesis for Pacha's film slate and badges. Use when the user asks for partners, outreach emails/DMs, community management, campaigns, content calendar, UGC rituals, or community signal for decisions—not for competitive film scoring (use pacha-producer-jury).

Turns Draft Editor work-log entries marked ingest=true into durable knowledge: append to a digest markdown file and re-run meta-layer graph ingest. Use for weekly sweeps, habit capture follow-up, or when the user says to process the work log / ingest links from the calendar.

Fetches a YouTube transcript via TranscriptAPI, writes `.txt` and `.json` under `data/transcripts/youtube/`, and never assumes a default video. Use when the user wants a YouTube transcript saved to disk, TranscriptAPI transcript export, or to avoid losing transcript output in the terminal. Requires an explicit URL or video id every time.

Captures a product, agent, or monument idea into the repo ideas backlog from the current chat. Use when the user asks to capture an idea, park an idea, idea backlog, brain dump, feature wishlist, "note this for later", or similar—without turning it into a full decision log entry.

Drafts a "conversation → build" log entry for agent-lab from the current chat: decision, context, files/commits, follow-ups. Use when the user asks to log a decision, draft a build log, record this conversation for the repo, update AGENT_AND_MONUMENT_DECISIONS, monument launch notes, or "what we talked about" for documentation.