merge: docs-v2-dev into docs-v2 (2026-04-07)

.claude/CLAUDE.md

@@ -39,10 +39,8 @@ Alison Haire (Wonderland). Documentation lead. Decision authority.
 | Snippets Audit | Full audit of `snippets/automations` and `snippets/data` with creator/consumer/risk report in `snippets/_workspace/reports/automations-data-audit-2026-04-05.md` | Done | 2026-04-05 |
 | Delegators | Canonical `v2/delegators` IA rebuild: live route migration, glossary move, reference pages, nav/redirect propagation, generated surface sync | Done | 2026-04-06 |
 | Gateways Verify | Consolidated verify page (`v2/gateways/setup/verify/test.mdx`) — health checks, end-to-end tests, on-chain verification for video/AI/dual across Docker/Linux/Windows. Research collated, 3 OS child views built. Scoped distinct from monitor (Prometheus/Grafana) | Done | 2026-04-07 |
-| Gateways Monitoring | Holistic `monitor.mdx` at `v2/gateways/setup/` with Docker/Linux/Windows tabs, Prometheus, Grafana, Explorer embed, alerts | Done | 2026-04-07 |
-| Gateways Prepare | Consolidated `prepare.mdx` at `v2/gateways/setup/` replacing Setup Checklist group. Off-chain/On-chain tabs, hardware/network/OS reqs, RPC/wallet/keystore setup | Done | 2026-04-07 |
-| Gateways Connect | Holistic `connect.mdx` at `v2/gateways/setup/` with Docker/Linux/Windows tabs. Off-chain/on-chain synced nested tabs, discovery verification, service publication, selection tuning. Research collated, 3 OS view files + parent page + nav registration | Done | 2026-04-07 |
-| Content Structure Templates | 3-layer page layout framework: classification guide, 13 composable blocks, 7 pageType templates, agent prompt. 24 files in `snippets/templates/`. Research + design + build + iterate complete. Next: test on real pages, fill golden examples for 6 missing types | Done | 2026-04-07 |
+| Gateways Monitoring | Holistic `monitor.mdx` for `v2/gateways/setup/monitor/` with Docker/Linux/Windows tabs, Prometheus, Grafana, Explorer, alerts. Research + OS view files + parent page created | Active - page built, awaiting review | 2026-04-07 |
+| Gateways Connect | Holistic `connect.mdx` at `v2/gateways/setup/` with Docker/Linux/Windows tabs. Off-chain/on-chain connection, discovery verification, service publication, selection tuning. Research collated, 3 OS view files + parent page + nav registration | Active - page built, testing | 2026-04-07 |
 
 **Rule:** When you finish a task or change status, update your row in this table before closing. If the master-tasks.md file has a matching item, update that too.
 
@@ -55,7 +53,7 @@ Alison Haire (Wonderland). Documentation lead. Decision authority.
 - **Reproduce before fixing.** For bugs: (1) reproduce the exact failure, (2) capture the error, (3) hypothesise from evidence, (4) test one fix at a time. Use `/diagnose` for anything beyond a trivial one-line fix.
 - **Verify first instance before bulk operations.** Apply to one file, confirm it works, then proceed to the rest.
 - **Read Mintlify constraints before editing MDX.** Reference: `docs-guide/canonical/collation-data/Mintlify/mintlify-repo-best-practices.md`. No React/hook imports, no Mintlify global imports, use root-absolute imports for shared resources, include file extensions, keep MDX-facing JSX data flow in parent MDX, define risky constants inside export bodies, use arrow function syntax only, and follow the repo's scoped preview and styling rules.
-- **Verify renders before declaring done.** After editing any `v2/*.mdx` file, confirm the PostToolUse render-verify hook reported PASSED. If it reported `server-failed`, restart the server with `node operations/scripts/dispatch/governance/server-lifecycle.js restart` before continuing. Before declaring a page complete, run the smoke test: `node operations/tests/integration/mdx-component-runtime-smoke.js --routes /v2/path/to/page`. Never use `node .githooks/server-manager.js` directly (library, not CLI). Never declare a page "done" without a PASSED verification or successful smoke test.
+- **Verify renders before declaring done.** After editing any `v2/*.mdx` file, confirm the PostToolUse render-verify hook reported PASSED. If it reported `server-failed`, restart the server with a scoped restart: `node operations/scripts/dispatch/governance/server-lifecycle.js restart v2/TAB` (e.g. `restart v2/home`). Scoped restarts take <2 min. Never cold-start the full 795-page docs.json (10+ min). Before declaring a page complete, run the smoke test: `node operations/tests/integration/mdx-component-runtime-smoke.js --routes /v2/path/to/page`. Never use `node .githooks/server-manager.js` directly (library, not CLI). Never declare a page "done" without a PASSED verification or successful smoke test.
 
 ---
 
@@ -165,7 +163,7 @@ These are inline. No skill files. Just do what it says.
 | `/remind` | Re-read `.claude/CLAUDE.md` right now. Re-read the thread outcome. State your role, the rules you're following, and what you're working on. Then propose the next action with reasoning. |
 | `/log` | Append a timestamped entry to `workspace/thread-outputs/sessions/session-log.txt` with: thread name, what just shipped, files changed. One entry, no ceremony. Then continue working. |
 | `/flag [thread] [message]` | Append a flag to `workspace/thread-outputs/sessions/flags.jsonl` for another thread to pick up. One JSON line. Then continue working. |
-| `/server` | Check server health: `node operations/scripts/dispatch/governance/server-lifecycle.js health`. If down, restart: `server-lifecycle.js restart`. Read log on failure: `tail -50 /tmp/mint-dev-test-*.log`. Report status. |
+| `/server` | Check health: `node server-lifecycle.js health`. If down, scoped restart: `node server-lifecycle.js restart v2/TAB` (e.g. `restart v2/home`). Read log on failure: `tail -50 /tmp/mint-dev-test-*.log`. Full path: `operations/scripts/dispatch/governance/server-lifecycle.js`. |
 
 **If the user says a message was lost/eaten:** Immediately read the last 5 entries from `workspace/thread-outputs/sessions/message-backup.jsonl` and present them. Don't ask questions — just show the content.
 

.claude/projects/-Users-alisonhaire-Documents-Livepeer-Docs-v2-dev/memory/MEMORY.md

@@ -0,0 +1,23 @@
+# Memory Index
+
+## CRITICAL
+- [Failure 1: Worktree only - no direct writes](feedback_critical_failure_worktree.md) - NEVER write to docs-v2-dev. Worktree only. NEVER destructive git. NEVER spawn agents without approval. Findings before fixes always. Highest priority rule.
+- [Failure 2: Repeated git checkout same session](feedback_critical_failure_2_same_session.md) - Ran git checkout HEAD -- on branch AGAIN 30 minutes after writing the rule. Overwrote files with old snapshots then destroyed more work. Advisory rules do not work.
+
+## Feedback (consolidated)
+- [Process discipline](feedback_process_discipline.md) - Contextual detail for each step of the CLAUDE.md reasoning loop: what to read, when to document, how to ground actions
+- [Speed and mechanics](feedback_speed_and_mechanics.md) - 5-second responses, agent delegation, direct file writes, focus protection, no redundant questions
+- [Communication style](feedback_communication_style.md) - No apologies, no em dashes, short headings, short and correct output, frustration is diagnostic
+- [Safety and scope](feedback_safety_and_scope.md) - Branch safety, task scope limits, destructive action evidence, root cause through generators, mint dev
+- [Content terminology](feedback_content_terminology.md) - On-chain/off-chain is operational mode; three independent axes (operational mode, setup type, node type)
+- [IA ownership](feedback_ia_ownership.md) - Alison designed the IA; AI runs are secondary sources. Only pull the best from AI output.
+- [Working style](feedback_working_style.md) - Carry the load, high standards, no drift. Alison course-corrects, not drives. Execute, don't discuss.
+- [Agent quality](feedback_agent_quality.md) - Every agent must self-test, reason from first principles, bring coordination/context/validation. No unchecked claims.
+- [No branch switching](feedback_no_branch_switching.md) - Never create new branches or switch away from current working branch. All work on docs-v2-dev.
+- [Agent workflow pattern](feedback_agent_workflow.md) - Coordinator model: record tasks, background agents do work + verify, never block chat. Auto-log to record-log.md.
+- [Parallel execution](feedback_parallel_execution.md) - Never queue independent tasks. If approach is agreed, run all in parallel immediately.
+- [Quality standard](feedback_quality_standard.md) - Verify before reporting done. Task list/record-log/standard are separate things. Mermaid types need different themes. Check preset defaults.
+
+## Project
+- [Style guide outdated](project_style_guide_outdated.md) - Style guide validator has 637 errors; needs its own governance thread (flagged 2026-03-20)
+- [Mintlify dev audit](project_mint_dev_audit.md) - Multi-session audit complete: 0 parsing errors, most warnings cleared. Three known-accepted warnings remain

.claude/projects/-Users-alisonhaire-Documents-Livepeer-Docs-v2-dev/memory/feedback_confirm_before_execute.md

@@ -0,0 +1,18 @@
+---
+name: confirm_before_execute
+description: Always confirm task scope and get explicit go-ahead before executing any implementation work
+type: feedback
+---
+
+**Rule: Confirm tasks before executing. Never implement without explicit approval.**
+
+When the user describes a problem, asks "what needs to be done", or says "we want to do X" — treat this as a request for a report or plan, NOT a go-ahead to execute.
+
+**Why:** User said "i didn't ask you to implement anything. i just wanted a fucking report" after I executed a large batch of v2 folder renames and policy changes without being asked. Direct quote: "fucking listen and confirm tasks FIRST".
+
+**How to apply:**
+- When a user describes a problem or names a task area → produce a report/analysis of current state and what would need to happen
+- Only execute when the user gives an explicit instruction: "do it", "implement this", "go ahead", "execute", etc.
+- If unsure whether they want a report or execution, ask: "Do you want me to do this, or just summarize what needs doing?"
+- "We want to do X" = they are describing intent, NOT giving a go-ahead
+- "Let's do X" without further context = ask before acting

.github/workflows/README-translate-docs.md

@@ -9,7 +9,7 @@ The workflow also fails non-dry runs if the checked-in provider is accidentally
 
 1. Installs `tools` dependencies
 2. Runs i18n unit tests (`npm --prefix tools run test:i18n`)
-3. Runs `operations/scripts/automations/content/language-translation/translate-docs.js`
+3. Runs `operations/scripts/i18n/translate-docs.js`
 4. Updates `docs.json` (`generate-localized-docs-json.js`) when not dry-run
 5. Validates generated localized MDX (`validate-generated.js`) when not dry-run
 6. Validates docs navigation (`tests/unit/docs-navigation.test.js`) when not dry-run

.github/workflows/broken-links.yml

@@ -21,7 +21,6 @@ jobs:
             v2
             snippets
             operations/scripts/snippets
-            operations/scripts/automations/content/data/fetching
           sparse-checkout-cone-mode: false
           fetch-depth: 0
 

.github/workflows/generate-ai-companions.yml

@@ -56,6 +56,6 @@ jobs:
         run: |
           git config user.name "GitHub Actions Bot"
           git config user.email "actions@github.com"
-          find v2 -type f -name '*-data.json' -exec git add -- {} +
+          git add v2/**/*-data.json
           git commit -m "chore(ai-companions): regenerate glossary companion JSONs - $(date -u +"%Y-%m-%dT%H:%M:%SZ")"
           git push origin HEAD:${{ steps.branch.outputs.name }}

.github/workflows/generate-llms-files.yml

@@ -43,10 +43,6 @@ jobs:
         run: |
           git config user.name "github-actions[bot]"
           git config user.email "github-actions[bot]@users.noreply.github.com"
-          for file in llms.txt llms-full.txt; do
-            if [ -e "$file" ] || git ls-files --error-unmatch "$file" >/dev/null 2>&1; then
-              git add "$file"
-            fi
-          done
+          git add llms.txt
           git diff --cached --quiet || git commit -m "chore: regenerate llms.txt [skip ci]"
           git push origin HEAD:${{ steps.branch.outputs.name }}

.github/workflows/update-livepeer-release.yml

@@ -32,5 +32,5 @@ jobs:
     with:
       mode: livepeer-release
       use_test_branch: ${{ github.event_name == 'workflow_dispatch' && inputs.use_test_branch || false }}
-      version: ${{ github.event_name == 'repository_dispatch' && github.event.client_payload.version || github.event_name == 'workflow_dispatch' && inputs.version || '' }}
+      version: ${{ github.event_name == 'repository_dispatch' && github.event.client_payload.version || inputs.version || '' }}
     secrets: inherit

.github/workspace/actions-library/generators/discoverability/generate-llms-files.mdx

@@ -60,7 +60,7 @@ graph LR
 ## Dependencies
 
 **Scripts:**
-- `operations/scripts/generators/ai/llm/generate-llms-files.js`
+- `operations/scripts/generate-llms-files.js`
 
 <CustomDivider />
 

.github/workspace/actions-library/generators/maintenance/generate-action-docs.yml

@@ -11,7 +11,7 @@ name: Generate Action Documentation
 on:
   push:
     branches:
-      - docs-v2
+      - ${{ vars.DEPLOY_BRANCH }}
     paths:
       - ".github/workflows/**"
       - ".github/workspace/actions-library/actions-audit.json"
@@ -35,9 +35,6 @@ jobs:
   generate-docs:
     runs-on: ubuntu-latest
 
-    env:
-      DRY_RUN: ${{ github.event_name == 'workflow_dispatch' && inputs.dry_run || false }}
-
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -54,7 +51,7 @@ jobs:
         run: |
           set -euo pipefail
           FLAGS=""
-          if [[ "$DRY_RUN" == "true" ]]; then FLAGS="--dry-run"; fi
+          if [[ "${{ inputs.dry_run }}" == "true" ]]; then FLAGS="--dry-run"; fi
           node .github/workspace/generate-action-pages.js $FLAGS
 
       - name: Check for changes
@@ -70,7 +67,7 @@ jobs:
           fi
 
       - name: Commit and push
-        if: steps.changes.outputs.changed == 'true' && env.DRY_RUN != 'true'
+        if: steps.changes.outputs.changed == 'true' && inputs.dry_run != 'true'
         run: |
           git config user.name "github-actions[bot]"
           git config user.email "github-actions[bot]@users.noreply.github.com"
@@ -87,6 +84,6 @@ jobs:
           else
             echo "No changes detected. Pages are up to date." >> $GITHUB_STEP_SUMMARY
           fi
-          if [[ "$DRY_RUN" == "true" ]]; then
+          if [[ "${{ inputs.dry_run }}" == "true" ]]; then
             echo "*(dry-run mode, no files written)*" >> $GITHUB_STEP_SUMMARY
           fi

.github/workspace/actions-library/generators/maintenance/sdk_generation.mdx

@@ -1,7 +1,7 @@
 ---
-title: 'Generate SDK Clients'
-sidebarTitle: 'Generate SDK Clients'
-description: 'Generate SDK Clients'
+title: 'Generate'
+sidebarTitle: 'Generate'
+description: 'Generate'
 pageType: reference
 purpose: reference
 audience: internal

.github/workspace/actions-library/notes

@@ -18,11 +18,11 @@ type-concern-function-<-script-name.yml>
 
 where:
 type: the category of script or action it belongs to
-= [automation, audit, dispatch, generate, remediate, validate, interface]
-concern: what part of the system the workflow operates on
-= [copy, governance, documentation]
+=[automation, audit, dispatch, generate, remediate, validate, interface]
+concern: what part of the system the workflow operates on.
+= [copy, governance, documentation, ]
 function: what this script does
-= [update, monitor]
+=[update, nonitor, ]
 
 eg. update-contract-addresses.yaml
 -> update is the function (we should enum these also)