메뉴
YouTube transcript ingestion pipeline and RAG query interface for the AI Engineer persona. All 763 @aiDotEngineer channel videos indexed in Qdrant. Run weekly to stay current.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Local-only multi-agent orchestrator with token optimization, 6-signal composite routing, and persistent memory. Routes all tasks through local executor bridges (Claude Code, Hermes, Gemini, Codex) with DAG dependencies, auto-episode creation, and cognitive profiles.
Zouroboros self-healing watchdog that monitors scheduled agent models and automatically switches failing agents to backup models, restoring originals when they recover.
Hybrid SQLite + Vector persona memory system for Zo Computer. Episodic memory with temporal queries, graph-boosted search, BFS path finding, knowledge gap analysis, auto-capture pipeline, and per-workload model routing. Default generation workloads use OpenAI `gpt-4o-mini`; embeddings use OpenAI `text-embedding-3-small`.
Security governance gate for AI agent skills, backed by NVIDIA SkillSpector. Statically scans a skill (local path, directory, zip, or Git URL) for 65 vulnerability patterns across 16 categories (prompt injection, data exfiltration, privilege escalation, supply chain, excessive agency, tool/MCP misuse, dangerous code via AST, YARA signatures) and returns a 0-100 risk score with a SAFE / REVIEW / DO_NOT_INSTALL recommendation. Use BEFORE adopting any third-party skill, and for the periodic baseline scan of the Skills/ directory.
Detector for model classifier SOFT-blocks — refusals, filtered-to-empty, truncated-with-disclaimer, content-policy flags — that masquerade as successful empty output to existing infrastructure (agent-model-healer only catches HARD failures: HTTP 4xx/5xx, timeouts, genuine empties). Classifies an output as soft_block / genuine_empty / ok using a per-provider signal catalog, then routes a per-domain fallback (security → broader-tolerance model, bio/chem → human review, distillation → refuse-by-design, never routed around). Every block + fallback is appended to an audit ledger. A quarterly system-card-diff agent flags expanded classifier scope so the catalog stays current.
Rendered-output verification station for the swarm post-flight eval. After a maker subagent produces a visual deliverable (UI/route/site), the station captures a full-page screenshot via agent-browser, then an independent verifier model (not equal to author, per aiewf P0-2) reads the screenshot against seed acceptance criteria + project DESIGN.md tokens + prior screenshots. Match marks the task complete; mismatch emits a structured visual diff for the maker's next iteration. The verifier is the exit condition, not the maker. Recurring visual failure modes graduate to instincts.yaml via the extract-patterns gate (ZOU-452).
SOC 직업 분류 기준
| name | ai-engineer-learning |
| description | YouTube transcript ingestion pipeline and RAG query interface for the AI Engineer persona. All 763 @aiDotEngineer channel videos indexed in Qdrant. Run weekly to stay current. |
| compatibility | Created for Zo Computer |
| metadata | {"author":"marlandoj.zo.computer"} |
Consumes, indexes, and queries the entire @aiDotEngineer YouTube channel — a leading source of AI engineering talks from Google DeepMind, Anthropic, OpenAI, Hugging Face, and more.
ai-engineer-videos @ 127.0.0.1:6333# Semantic search (primary interface)
bun Skills/ai-engineer-learning/scripts/query.ts "how to build durable agents"
bun Skills/ai-engineer-learning/scripts/query.ts "context management" --top 10
bun Skills/ai-engineer-learning/scripts/query.ts "evaluation frameworks"
# Within AI Engineer persona, this is the first thing to call before suggesting approaches
# 1. Update channel-videos.txt with new videos
cd /home/workspace/Skills/ai-engineer-learning/scripts
bun update-channel.ts # or manually append to channel-videos.txt
# 2. Index new videos (metadata-only first — always works, no YouTube access needed)
python3 ingest-metadata.py
# 3. Upgrade to transcript-backed if transcripts are available via save_webpage
# (saves markdown to /home/workspace/Articles/, then process with:)
bun process-transcripts.ts
# Videos saved via Zo's save_webpage land in /home/workspace/Articles/
# as "* :: www.youtube.com.md" files. Run this to embed them:
bun Skills/ai-engineer-learning/scripts/process-transcripts.ts
Use this for natural language deep-dives on topics. Requires valid NotebookLM cookies.
# Check session status
cd /home/workspace/Skills/notebooklm-skill/scripts
python3.12 nlm.py status
# If expired — re-export cookies (90-second process):
# 1. Open notebooklm.google.com in Chrome (must be signed in)
# 2. Install Cookie-Editor extension, click Export, copy JSON
# 3. Paste to /tmp/cookies.json on Zo
# 4. python3.12 nlm.py import-cookies /tmp/cookies.json
# Bulk-add remaining 535 videos to 2 NotebookLM notebooks (one-shot after auth):
python3.12 /home/workspace/Skills/ai-engineer-learning/scripts/load-to-notebooklm.py
# Query NotebookLM knowledge base:
python3.12 nlm.py ask --notebook <nb_id> --question "best practices for agent memory"
# (Notebook IDs saved to Projects/ai-engineer-learning/notebooklm-notebooks.json after load)
scripts/ingest-metadata.py — Metadata-only Qdrant indexer (no YouTube access needed)scripts/ingest-batch.py — Full transcript Qdrant indexer (requires residential IP / save_webpage)scripts/process-transcripts.ts — Embeds Articles/ markdown saves into Qdrantscripts/load-to-notebooklm.py — Bulk-add YouTube URLs to NotebookLM notebooksscripts/query.ts — Semantic search + research notebook generatorProjects/ai-engineer-learning/channel-videos.txt — Video metadata catalog (763 videos)Projects/ai-engineer-learning/processed.json — State file (tracks ingested videos)Projects/ai-engineer-learning/notebooklm-notebooks.json — NotebookLM notebook IDs (after load)Weekly agent [ZBR] Ingest AI Engineer Videos runs every Monday at 0200 UTC:
source /root/.zo_secrets && python3 /home/workspace/Skills/ai-engineer-learning/scripts/ingest-metadata.py
This indexes new videos from channel-videos.txt and embeds them into Qdrant.