docs(user-docs): video library + per-page links, v0.6.1 What's New, sync dev features

[vybe]

Summary

Two things, landed together against dev:

1. Video enrichment — a new docs/user-docs/videos.md library (35 published videos, grouped by topic, newest-first) plus 📺 Watch callouts on 32 feature pages and a README "Watch" section. Each callout links the most relevant videos, prioritizing newer ones.
2. Doc sync — the user docs had drifted behind dev. This adds the missing user-facing whats-new/v0.6.1.md and documents the dev-only features that shipped since the last sync.

New pages

• agents/agent-runtimes.md — Claude Code / OpenAI Codex / Gemini CLI runtimes (Codex harness MVP — pluggable agentic execution engine alongside Claude Code #1187)
• agents/agent-data.md — data_paths + on-demand export/import portability (Agent runtime data volumes: declared data paths with snapshot/restore and portable export #1169)
• whats-new/v0.6.1.md — user-facing release highlights (no issue numbers/codenames)
• videos.md — the video library

Updated pages

• creating-agents (compatibility validation + auto-fix, feat: Agent deployment compatibility validation — server-side checks with auto-fix offers #668), help (in-app bug reporting, feat(ui): In-app bug reporting from the floating Help widget (hosted intake → GitHub issues) #1116), subscription-credentials (hot-reload rotation, feat: credential rotation via hot-reload, not container recreate #1089), agent-network (experimental pull-pilot routing, experiment: pull-coordination pilot — route MCP chat_with_agent through the agent queue #946), agent-configuration (AGENT_TMP_SIZE, fix: agent /tmp tmpfs (100 MB) fills and silently breaks autonomous git commits — make size configurable #1231), single-server (Postgres migration-runner groundwork — not yet a user-selectable backend, fix(db): migration runner — DROP-rebuild data-loss window and no cross-process serialization #1160), overview + setup (multi-runtime framing, setup-token notes), README index (+ the previously-orphaned agent-session).

Notes

• Deliberately not documented (no user surface): agent-server auth hardening (security: agent-server HTTP API is unauthenticated on the shared agent network (cross-agent credential theft) #1159), status-as-projection refactor (refactor: status-as-projection — schedule_executions.status never read as authoritative for "is running" #1082). TOTP login (feat(auth): two-factor (TOTP) login seam + enterprise-gated UI (trinity-enterprise#5) #1247) left for when the enterprise UI ships publicly.
• Verified: v0.6.1 carries no issue numbers; deploy guides pass the public-safety scan; all new-page links resolve; no secrets/PII/internal IPs in the diff.

🤖 Generated with Claude Code

[dolho]

/review Report — automated structural review

Branch: feature/user-docs-video-and-dev-sync → dev
Files: 41 (+414/−8) · Type: user docs · Scope: CLEAN

Critical Findings

None. All 41 files are .md under docs/ (no code, no config). Secret/PII sweep clean. External links resolve to youtu.be/*, github.com/abilityai/trinity, ability.ai/workshops, and localhost examples — no internal hosts or credentials leaked.

Informational Findings

[I1] Link liveness unverifiable in-CI (Confidence: 6/10)
~20 youtu.be/<id> video links added. A public/private/unlisted YouTube video still returns HTTP 200, so an automated HEAD check can't confirm they actually play — and a private/wrong ID would silently 404 for users.
Suggestion: spot-check the new video IDs render (or run a markdown link-checker that follows to the oEmbed endpoint, which does distinguish private videos). Same for the new per-page relative links — a lychee/markdown-link-check pass over docs/user-docs/ would catch dangling internal anchors.

[I2] Content accuracy not machine-verifiable (Confidence: 5/10)
"v0.6.1 What's New / sync dev features" describes shipped behavior; correctness depends on matching the actual dev feature set. Worth a human spot-read against the release notes, but nothing here looks stale.

Summary

• Critical: 0
• Informational: 2 — link-checker pass + a content spot-read recommended; neither blocks.
• Scope: CLEAN (pure user docs)

Automated /review.