titan-agent 5.4.2 → 5.5.6

package/docs/superpowers/plans/2026-04-29-comprehensive-test-plan.md

@@ -0,0 +1,396 @@
+# TITAN Comprehensive Test & Audit Plan
+
+**Goal:** Exhaustively test every TITAN subsystem to surface latent bugs, performance issues, configuration gaps, and user-facing regressions.
+
+**Approach:** Horizontal sweep across all layers (unit → integration → e2e → production). Each test bucket is independent and can run in parallel.
+
+---
+
+## Bucket 1: Core Agent Loop
+
+### 1.1 Think/Respond Cycle Invariants
+- [ ] Non-autonomous mode completes multi-step tasks (write → read → edit chain)
+- [ ] Autonomous mode respects maxRounds / maxToolRoundsHard limits
+- [ ] Voice fastPath caps tokens at 2000 and returns within 5s
+- [ ] Empty responses trigger ToolRescue (extractToolCallFromContent)
+- [ ] `no_tool_cap` stall detection fires after 2 nudges
+- [ ] `silence` stall detection fires after 120s inactivity
+- [ ] Budget warnings injected once per loop (not spam)
+
+### 1.2 Session Lifecycle
+- [ ] Session compaction triggers after maxSessionRuns / maxSessionAgeHours
+- [ ] Session timeout kills process after 30 minutes idle
+- [ ] Absolute timeout kills process after 10 minutes total
+- [ ] CleanupBudget removes spend tracking on session end
+- [ ] Steer queue cleared on session end
+
+### 1.3 Context Budgeting
+- [ ] `maxTokens=200000` never triggers false "budget exceeded" for short tasks
+- [ ] `maxTokens=4000` aggressively compresses to fit
+- [ ] `compress` action truncates oldest history first
+- [ ] `downgrade` action switches to cheap model at 80% spend
+- [ ] `stop` action hard-halts at 100% spend
+
+---
+
+## Bucket 2: Provider Layer
+
+### 2.1 Model Routing
+- [ ] `kimi-k2.6:cloud` routes to Ollama on Titan PC
+- [ ] `fast` alias resolves to `qwen3.5:cloud`
+- [ ] `smart` alias resolves to `glm-5:cloud`
+- [ ] Fallback chain activates on 503/504/timeout
+- [ ] `toolCapableModels` list gates tool injection
+- [ ] Auxiliary model (`minimax-m2.7:cloud`) used for goal proposing
+
+### 2.2 Provider Health
+- [ ] Ollama health check via `/api/tags` (not just TCP)
+- [ ] OpenRouter credit check before routing
+- [ ] Claude Code adapter handles `think` blocks correctly
+- [ ] Google Gemini handles `maxOutputTokens` vs `max_tokens` schema
+- [ ] `clampMaxTokens()` per model prevents 400 errors
+
+### 2.3 Rate Limiting
+- [ ] 429 from Ollama triggers 1s/2s/4s retry backoff
+- [ ] Credential rotation after cooldown period
+- [ ] Cost estimator warns before expensive call
+
+---
+
+## Bucket 3: Tool Execution
+
+### 3.1 Shell Tool
+- [ ] Blocked commands (rm -rf /, chmod 777 /etc, etc.) rejected
+- [ ] Command scanner scores commands 0-100 and blocks > threshold
+- [ ] Timeout default is 60000ms (not 30000ms)
+- [ ] Background mode (`background: true`) spawns without waiting
+- [ ] `verify_port` waits up to 20s for port to open
+- [ ] 10MB output buffer, truncates to 25KB head + 25KB tail
+
+### 3.2 File Tools
+- [ ] `write_file` creates directories recursively
+- [ ] `edit_file` produces unified diff in response
+- [ ] `apply_patch` handles fuzzy matching
+- [ ] Pre-execution snapshot captured for diff generation
+- [ ] File system allowlist blocks `.ssh`, `.env`, `.git/objects`
+
+### 3.3 Browser Automation
+- [ ] `browser_navigate` loads page with 60s timeout
+- [ ] `browser_screenshot` captures full page
+- [ ] `browser_auto_nav` handles CAPTCHA via capsolver
+- [ ] `web_act` executes click/type actions
+- [ ] Browser pool reuses sessions across calls
+
+### 3.4 Web Tools
+- [ ] `web_fetch` returns text content (not raw HTML)
+- [ ] `web_search` queries multiple engines and dedupes
+- [ ] `smart_form_fill` extracts form schema and fills fields
+- [ ] AbortSignal.timeout(45000) on all web calls
+
+### 3.5 Approval Gates
+- [ ] `requiresApproval()` checks per-tool preferences
+- [ ] `approvalPending` returned when unconfigured
+- [ ] Auto-approve for `safe` tools (read_file, list_dir)
+- [ ] Auto-deny for `dangerous` tools (shell, exec, code_exec)
+- [ ] Command-post approval requests show diff/context
+
+---
+
+## Bucket 4: Memory & Graph
+
+### 4.1 Episodic Memory
+- [ ] `graph_remember` stores episode with embedding
+- [ ] `graph_search` returns relevant episodes by vector similarity
+- [ ] `graph_entities` extracts people, places, concepts
+- [ ] `graph_recall` reconstructs timeline
+- [ ] Auto-compact when > maxHistoryMessages (50)
+
+### 4.2 Vector Search
+- [ ] `nomic-embed-text` model loaded for embeddings
+- [ ] 4290 vectors indexed at startup (Titan PC)
+- [ ] Query embedding matches within 512 tokens
+- [ ] Dimension check: 768d vectors
+
+### 4.3 Working Memory
+- [ ] Tracks session state (activeTools, summary, goals)
+- [ ] Retires old entries after pressure threshold
+- [ ] Persists to `~/.titan/working-memory.jsonl`
+
+---
+
+## Bucket 5: Self-Awareness & Autonomy
+
+### 5.1 Drive System
+- [ ] `curiosity`, `competence`, `connection`, `pressure` drives fire
+- [ ] Drive satisfaction score updates per tick
+- [ ] Goal proposals filed when drive satisfaction < threshold
+- [ ] `peerAdvise` gets verdict from sage sub-agent
+
+### 5.2 Self-Healing
+- [ ] `self_doctor` detects anomalies in logs
+- [ ] `self_repair` generates fix proposals
+- [ ] Proposals go to command post for approval
+- [ ] Circuit breaker trips after 50 rounds
+
+### 5.3 Organism
+- [ ] Hormones injected into system prompt
+- [ ] Pressure threshold (1.2) triggers organism response
+- [ ] Shadow mode runs background analysis
+- [ ] Tick interval 60s, not blocking main loop
+
+---
+
+## Bucket 6: Gateway & Server
+
+### 6.1 HTTP Endpoints
+- [ ] GET `/api/health` returns `{status: "ok", version, uptime}`
+- [ ] GET `/api/config` returns safe config (passwords masked)
+- [ ] GET `/api/stats` returns token usage, memory, provider health
+- [ ] POST `/api/message` accepts message + systemPromptAppendix
+- [ ] POST `/api/eval/run` runs eval suites
+
+### 6.2 Authentication
+- [ ] Password mode (`titan2026`) issues Bearer token
+- [ ] Token mode validates `Authorization: Bearer <token>`
+- [ ] `mode: "none"` bypasses auth entirely
+- [ ] Mesh auth validates secret + nodeId
+
+### 6.3 HTTPS
+- [ ] Auto-detects `~/.titan/certs/titan.pem` + `titan-key.pem`
+- [ ] Falls back to HTTP if certs missing
+- [ ] Health check responds to HTTPS (kimi-k2.6:cloud verified)
+
+### 6.4 WebSocket
+- [ ] `/ws` accepts agent bus connections
+- [ ] Mesh peers authenticate via URL params
+- [ ] Heartbeat every 60s
+
+### 6.5 SSE
+- [ ] `/api/stream` streams tool results in real-time
+- [ ] Reconnection handled on client disconnect
+
+---
+
+## Bucket 7: Channels
+
+### 7.1 WebChat
+- [ ] Enabled by default, runs on port 48421
+- [ ] Receives messages via POST `/api/webchat/message`
+- [ ] Returns widget HTML or text
+
+### 7.2 Messenger
+- [ ] Webhook at `/api/messenger/webhook`
+- [ ] Voice replies via F5-TTS (voice=andrew)
+- [ ] Page ID configured
+
+### 7.3 Twilio
+- [ ] Inbound calls trigger voice agent
+- [ ] TTS engine: f5-tts
+- [ ] Allowed callers list enforced
+
+### 7.4 Disabled Channels
+- [ ] Discord, Telegram, Slack, WhatsApp all `enabled: false`
+- [ ] No startup errors when disabled
+
+---
+
+## Bucket 8: Skills Registry
+
+### 8.1 Discovery
+- [ ] 248 tools loaded at startup (Titan PC log verified)
+- [ ] Auto-discovery scans `src/skills/builtin/` and `~/.titan/skills/`
+- [ ] Marketplace disabled (`marketplace: false`)
+
+### 8.2 Registration
+- [ ] Each skill has `name`, `description`, `parameters`, `execute`
+- [ ] Schema validation via Zod
+- [ ] Duplicate names rejected
+
+### 8.3 Procedural Memory
+- [ ] `save_skill` stores custom skills
+- [ ] `recall_skill` retrieves by fuzzy name match
+- [ ] `auto_generate_skill` creates from natural language
+
+---
+
+## Bucket 9: Sub-agents & Delegation
+
+### 9.1 Spawn
+- [ ] `spawn_agent` creates new process with isolated context
+- [ ] Max depth = 2 (grandchildren allowed, great-grandchildren blocked)
+- [ ] Worktree support for parallel edits
+
+### 9.2 Concurrency
+- [ ] `maxConcurrent: 8` sub-agents
+- [ ] `maxRoundsPerAgent: 10` per delegate
+- [ ] Pool queue when limit exceeded
+
+### 9.3 Results
+- [ ] Sub-agent results bubble back to parent
+- [ ] ` StructuredSpawn` formats output for parent consumption
+- [ ] Failed sub-agent returns error + partial output
+
+---
+
+## Bucket 10: Mesh / Networking
+
+### 10.1 Discovery
+- [ ] mDNS broadcasts on local network (`mdns: true`)
+- [ ] Tailscale VPN peers auto-discovered (`tailscale: true`)
+- [ ] Static peers from config (`staticPeers: []`)
+
+### 10.2 Communication
+- [ ] `allowRemoteModels: true` — peers can use our models
+- [ ] `maxRemoteTasks: 3` — queue limit
+- [ ] `heartbeatIntervalMs: 60000`
+- [ ] `peerStaleTimeoutMs: 300000` (5 min)
+
+### 10.3 Auth
+- [ ] `mesh.secret` required for peer connections
+- [ ] `autoApprove: false` — manual approval for mesh tasks
+
+---
+
+## Bucket 11: Voice
+
+### 11.1 TTS
+- [ ] F5-TTS server running on port 5006
+- [ ] GPU acceleration active
+- [ ] Voice `andrew` available
+- [ ] Max chars: 1000 per synthesis
+
+### 11.2 STT
+- [ ] faster-whisper endpoint: `http://localhost:48421`
+- [ ] LiveKit WebSocket: `ws://localhost:7880`
+
+---
+
+## Bucket 12: Safety & Security
+
+### 12.1 Kill Switch
+- [ ] `killSwitch.ts` monitors for dangerous patterns
+- [ ] Immediate halt on command injection detection
+- [ ] Alert sent to command post
+
+### 12.2 Secret Scanning
+- [ ] `secretGuard.ts` checks for API keys in output
+- [ ] Redacts passwords, tokens, private keys
+- [ ] Configurable deny list
+
+### 12.3 Command Scanner
+- [ ] Risk score 0-100 per command
+- [ ] Block level: commands rejected
+- [ ] Warn level: logged but allowed
+
+### 12.4 Sandbox
+- [ ] Docker sandbox: `image: titan-sandbox`
+- [ ] Openshell sandbox: NVIDIA GPU passthrough
+- [ ] Memory limit: 8192 MB, CPU: 1
+
+---
+
+## Bucket 13: VRAM & Performance
+
+### 13.1 GPU Monitoring
+- [ ] Ollama VRAM usage tracked every 10s
+- [ ] `autoSwapModel: true` when VRAM < reserveMB (3260)
+- [ ] Fallback to `qwen3:7b` on OOM
+
+### 13.2 Memory Usage
+- [ ] Gateway RSS: ~217 MB (Titan PC verified)
+- [ ] Heap used: ~81 MB
+- [ ] No memory leaks over 7-day uptime
+
+### 13.3 Response Times
+- [ ] Fast model (qwen3.5): < 5s
+- [ ] Smart model (glm-5): < 15s
+- [ ] Reasoning model (kimi-k2.6): < 60s for complex tasks
+- [ ] Ollama 503 retries: max 3 attempts with backoff
+
+---
+
+## Bucket 14: Configuration
+
+### 14.1 Schema Validation
+- [ ] `zod` schema catches invalid config on load
+- [ ] Defaults applied for missing fields
+- [ ] Type coercion for numeric strings
+
+### 14.2 Hot Reload
+- [ ] Config changes trigger component restart
+- [ ] `RESTART_REQUIRED_PATTERNS` list accurate
+- [ ] Graceful shutdown: finish in-flight requests
+
+### 14.3 Environment Variables
+- [ ] `TITAN_HOME` overrides default `~/.titan`
+- [ ] `TITAN_TELEMETRY_ENABLED=false` disables PostHog
+- [ ] `NODE_OPTIONS=--max-old-space-size=6144` for CI
+
+---
+
+## Bucket 15: CI/CD
+
+### 15.1 GitHub Actions
+- [ ] `ci.yml`: build + test on Node 22
+- [ ] `eval-gate.yml`: boot gateway + run eval suites
+- [ ] Gateway boot timeout: 180s (was 60s)
+- [ ] Artifact upload: eval results + gateway log retained 30 days
+
+### 15.2 Test Suites
+- [ ] Unit tests: `npm test -- --run`
+- [ ] Eval suites: 11 behavioral test suites
+- [ ] Threshold: 80% pass rate per suite
+- [ ] Sharding: 4 shards for memory headroom
+
+### 15.3npm Publish
+- [ ] Version bump: `package.json` + `constants.ts` + `README.md`
+- [ ] `npm publish --tag latest`
+- [ ] `npm dist-tag ls titan-agent` shows `latest: 5.4.2`
+
+---
+
+## Bucket 16: Updater
+
+### 16.1 Development Mode
+- [ ] `.git` detected → `git pull && npm run build`
+- [ ] Restart script spawns detached process
+
+### 16.2 Production Mode (systemd)
+- [ ] `/run/systemd/system` detected → use systemctl
+- [ ] `git pull && npm run build`
+- [ ] `sudo systemctl restart titan-gateway`
+
+### 16.3 Global npm Mode
+- [ ] `npm update -g titan-agent`
+- [ ] Requires write access to npm prefix
+
+---
+
+## Bucket 17: Telemetry & Analytics
+
+### 17.1 PostHog
+- [ ] API key: `phc_kVw5xLJx5SVXex9RSTCFwP8cJSNEXTYZ7oJwqoDdMPJX`
+- [ ] Host: `https://us.i.posthog.com`
+- [ ] Consent version: `5.0.1`
+
+### 17.2 Local Logging
+- [ ] `~/.titan/telemetry-events.jsonl`
+- [ ] `~/.titan/bug-reports.jsonl`
+- [ ] `~/.titan/activity-log.jsonl`
+
+---
+
+## Execution Plan
+
+Run all tests in parallel where possible:
+
+```bash
+# Run locally (Titan PC)
+ssh titan "cd /opt/TITAN && npm test -- --run 2>&1 | tail -20"
+
+# Run eval suites
+ssh titan "cd /opt/TITAN && node -e \"require('./dist/eval/harness.js').runAllSuites().then(console.log)\""
+
+# Gateway health soak test
+watch -n 5 'curl -k -s -H "Authorization: Bearer $(curl -k -s -X POST https://127.0.0.1:48420/api/login -H "Content-Type: application/json" -d "{\"password\":\"titan2026\"}" | python3 -c "import sys,json; print(json.load(sys.stdin)[\"token\"])")" https://127.0.0.1:48420/api/health'
+```

package/docs/superpowers/plans/2026-04-29-fix-all-prs.md

@@ -0,0 +1,251 @@
+# Fix All PRs + Production Issues Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:executing-plans
+
+**Goal:** Resolve all 3 failing GitHub PRs (dependabot deps, widget gates, docker-compose) + fix production gateway issues on Titan PC + publish v5.4.2.
+
+**Architecture:** Merge CI timeout fix (180s) into all PR branches so GitHub Actions can boot the gateway. Resolve conflicts in docker-compose PR by preferring PR changes. Verify Titan PC gateway health with HTTPS auth.
+
+**Tech Stack:** Git, GitHub CLI (`gh`), Node.js, npm, systemd, ssh
+
+---
+
+## Task 1: Verify dependabot PR #65 CI Status
+
+**Files:** None (just verification)
+
+- [ ] **Step 1: Check if PR #65 CI eval-gate passed**
+
+```bash
+ssh titan "cd /tmp/titan-full && gh run list --repo Djtony707/TITAN --branch dependabot/npm_and_yarn/production-deps-ec2688f007 --limit 3"
+```
+
+Expected: New `in_progress` run from re-open + merge, should eventually show `completed`.
+
+- [ ] **Step 2: If passed, merge PR #65**
+
+```bash
+ssh titan "cd /tmp/titan-full && gh pr merge 65 --repo Djtony707/TITAN --merge 2>&1"
+```
+
+Expected: `✓ Pull request #65 was merged.`
+
+---
+
+## Task 2: Merge main fix into widget-gates PR #64
+
+**Files:** `.github/workflows/eval-gate.yml` (from main)
+
+- [ ] **Step 1: Clone and prepare branch**
+
+```bash
+ssh titan "rm -rf /tmp/titan-fix64 && cd /tmp && git clone --depth 50 https://github.com/Djtony707/TITAN.git titan-fix64 && cd titan-fix64 && git fetch origin feat/group-a-fixes && git checkout -B feat/group-a-fixes origin/feat/group-a-fixes"
+```
+
+- [ ] **Step 2: Merge main into branch**
+
+```bash
+ssh titan "cd /tmp/titan-fix64 && git merge origin/main --no-edit 2>&1 || echo 'CONFLICT!'
+```
+
+Expected: `Already up to date.` OR clean merge.
+
+- [ ] **Step 3: Push and verify CI**
+
+```bash
+ssh titan "cd /tmp/titan-fix64 && git push origin feat/group-a-fixes 2>&1"
+```
+
+Expected: push successful, GitHub CI auto-triggers.
+
+- [ ] **Step 4: Wait for CI and merge**
+
+```bash
+sleep 30
+ssh titan "cd /tmp/titan-fix64 && gh pr checks 64 --repo Djtony707/TITAN"
+```
+
+Expected: all checks pass. Then:
+
+```bash
+ssh titan "cd /tmp/titan-fix64 && gh pr merge 64 --repo Djtony707/TITAN --merge 2>&1"
+```
+
+---
+
+## Task 3: Resolve conflicts in docker-compose-revert PR #63
+
+**Files:**
+- Modify: `docker-compose.yml` (keep PR version — it reverts to canonical)
+- Modify: `tests/gateway-extended.test.ts`, `tests/mesh-extended.test.ts`, `tests/ollama.test.ts`, `tests/providers-extended.test.ts`, `tests/subAgent.test.ts` (accept PR or main — whichever is newer)
+
+- [ ] **Step 1: Clone and prepare branch**
+
+```bash
+ssh titan "rm -rf /tmp/titan-fix63 && cd /tmp && git clone --depth 50 https://github.com/Djtony707/TITAN.git titan-fix63 && cd titan-fix63 && git fetch origin fix/docker-compose-revert && git checkout -B fix/docker-compose-revert origin/fix/docker-compose-revert"
+```
+
+- [ ] **Step 2: Start merge of main**
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && git merge origin/main --no-edit 2>&1 || echo 'CONFLICT_EXPECTED'"
+```
+
+Expected: CONFLICT in `docker-compose.yml` and several test files.
+
+- [ ] **Step 3: Resolve docker-compose.yml**
+PR #63 title says "fix(deploy): revert docker-compose.yml to canonical TITAN config". Keep **PR's version** of `docker-compose.yml`.
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && git checkout --ours docker-compose.yml && git add docker-compose.yml"
+```
+
+- [ ] **Step 4: Resolve test files**
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && for f in tests/gateway-extended.test.ts tests/mesh-extended.test.ts tests/ollama.test.ts tests/providers-extended.test.ts tests/subAgent.test.ts; do git checkout --ours \$f && git add \$f; done"
+```
+
+- [ ] **Step 5: Complete merge**
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && git status --short && echo '---'"
+```
+
+Should show only merge commit pending.
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && git commit --no-edit"
+```
+
+- [ ] **Step 6: Push and merge**
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && git push origin fix/docker-compose-revert"
+```
+
+Wait for CI:
+
+```bash
+sleep 30
+ssh titan "cd /tmp/titan-fix63 && gh pr checks 63 --repo Djtony707/TITAN"
+```
+
+Then merge:
+
+```bash
+ssh titan "cd /tmp/titan-fix63 && gh pr merge 63 --repo Djtony707/TITAN --merge 2>&1"
+```
+
+---
+
+## Task 4: Bump version + publish v5.4.2 to npm
+
+**Files:**
+- Modify: `package.json` (version already 5.4.2)
+- Modify: `src/utils/constants.ts` (TITAN_VERSION already 5.4.2)
+- Modify: `README.md` (version already 5.4.2)
+
+- [ ] **Step 1: Verify all version strings**
+
+```bash
+grep -r "5\.4\.2" package.json src/utils/constants.ts README.md
+```
+
+Expected: All show 5.4.2.
+
+- [ ] **Step 2: Build and test locally**
+
+```bash
+cd /Users/michaelelliott/Desktop/TitanBot/TITAN-main && npm run build && npm test -- --run 2>&1 | tail -5
+```
+
+Expected: 0 failing tests (or only pre-existing failures).
+
+- [ ] **Step 3: Publish to npm**
+
+```bash
+cd /Users/michaelelliott/Desktop/TitanBot/TITAN-main && npm publish --tag latest
+```
+
+Expected: `+ titan-agent@5.4.2`
+
+- [ ] **Step 4: Verify**
+
+```bash
+npm dist-tag ls titan-agent
+```
+
+Expected: `latest: 5.4.2`
+
+---
+
+## Task 5: Titan PC final verification
+
+**Files:** None (verification only)
+
+- [ ] **Step 1: Confirm gateway status**
+
+```bash
+ssh titan "systemctl is-active titan-gateway && curl -k -s http://127.0.0.1:48420/api/health --max-time 5"
+```
+
+Expected: `active` + `{"status":"ok",...}`
+
+- [ ] **Step 2: Update Titan PC config (already done)**
+
+```bash
+ssh titan "jq '.agent.maxTokens' /home/dj/.titan/titan.json"
+```
+
+Expected: `200000`
+
+- [ ] **Step 3: Pull latest on Titan PC**
+
+```bash
+ssh titan "cd /opt/TITAN && git pull origin main --ff-only && echo 'UP_TO_DATE'"
+```
+
+Expected: Already up to date (or fast-forward if new commits).
+
+---
+
+## Task 6: Update documentation / handoff
+
+**Files:**
+- Modify: `docs/HANDOFF/2026-04-29.md`
+
+- [ ] **Step 1: Append resolution notes**
+
+Append to `docs/HANDOFF/2026-04-29.md`:
+
+```markdown
+## Fixes Applied (continued)
+
+- PR #65 (dependabot): Merged CI timeout fix → main + branch
+- PR #64 (widget gates): Merged main → branch, CI should pass now
+- PR #63 (docker-compose): Resolved merge conflicts, kept canonical docker-compose.yml
+- npm: Published titan-agent@5.4.2
+- Titan PC: Token budget 130K→200K, gateway healthy
+```
+
+- [ ] **Step 2: Commit and push**
+
+```bash
+cd /Users/michaelelliott/Desktop/TitanBot/TITAN-main && git add docs/HANDOFF/2026-04-29.md && git commit -m "docs: PR resolution + npm publish handoff" && git push origin main
+```
+
+---
+
+## Verification Table
+
+| Item | Status | Check Command |
+|------|--------|---------------|
+| PR #65 merged | ⏳ `gh pr view 65` |
+| PR #64 merged | ⏳ `gh pr view 64` |
+| PR #63 merged | ⏳ `gh pr view 63` |
+| npm @latest | ⏳ `npm dist-tag ls titan-agent` |
+| Titan PC token budget | ✅ 200000 |
+| Titan PC gateway | ✅ `systemctl is-active titan-gateway` |
+| Version strings | ✅ 5.4.2 everywhere |
+