@@ -5,7 +5,7 @@
 # Awesome Hermes Agent Use Cases
 
 [![Awesome](https://awesome.re/badge.svg)](https://awesome.re)
-![Use Cases](https://img.shields.io/badge/usecases-55-blue?style=flat-square)
+![Use Cases](https://img.shields.io/badge/usecases-67-blue?style=flat-square)
 ![Last Update](https://img.shields.io/github/last-commit/ali-erfan-dev/awesome-hermes-usecases?label=Last%20Update&style=flat-square)
 ![Stars](https://img.shields.io/github/stars/ali-erfan-dev/awesome-hermes-usecases?style=flat-square)
 ![License](https://img.shields.io/badge/license-MIT-green?style=flat-square)
@@ -59,6 +59,8 @@ Then pick a use case below. The [`/demos`](demos/) folder has three runnable sta
 | [X Social Automation Toolset](usecases/x-social-automation-toolset.md) | A native Hermes plugin for X/Twitter search, account reads, trend checks, and operator-approved actions through Xquik. |
 | [Inbox → Slack Daily Digest](usecases/inbox-slack-daily-digest.md) | A weekday 9am cron that classifies your inbox and posts a needs-reply / FYI digest to a Slack channel. No auto-replies. |
 | [Sequenzy Email Marketing Agent](usecases/sequenzy-email-marketing-agent.md) | Use Hermes with Sequenzy skills/CLI to draft lifecycle campaigns, manage subscriber workflows, send tests, and monitor email stats. |
+| [Microsoft Teams Meeting → Action Pipeline](usecases/teams-meeting-action-pipeline.md) | Ingest Microsoft Graph meeting events, prefer transcripts, fall back to recording plus STT, and deliver durable summaries to Teams, Notion, or Linear. |
+| [Self-Hosted Nextcloud Workspace Assistant](usecases/nextcloud-workspace-assistant.md) | Manage Nextcloud files, notes, calendars, tasks, and contacts from Hermes through WebDAV, CalDAV, CardDAV, and the Notes API. |
 
 ## Messaging & Team Bots
 
@@ -117,26 +119,34 @@ Then pick a use case below. The [`/demos`](demos/) folder has three runnable sta
 | [Dynamic Skill Dojo](usecases/dynamic-skill-dojo.md) | A monitoring and improvement loop that reads session history, ranks weak skills, proposes patches or new skills, invokes self-evolution, and reports learning progress. |
 | [Hermes Kanban: Multi-Agent Swarm from Slack](usecases/kanban-multi-agent-swarm.md) | The v0.15 `/kanban` slash command turns a Slack message into a durable multi-agent swarm with named workers, a verifier, and a synthesizer — persistent via `~/.hermes/kanban.db`. |
 | [Deterministic Workflow Delegation](usecases/deterministic-workflow-delegation.md) | Hermes keeps the judgment, a deterministic runner keeps the receipts: repeatable jobs graduate from chat into checked `.nika.yaml` DAG files — audited before any token is spent, budget-capped, with tamper-evident run traces. Offline demo, zero keys. |
+| [Native A2A Agent Collaboration](usecases/native-a2a-agent-collaboration.md) | Discover and call specialist agents across machines or frameworks, or expose Hermes as an authenticated A2A service with streaming, audit logs, and anti-loop controls. |
+| [Per-Project Scoped Hermes Agents](usecases/per-project-scoped-hermes-agents.md) | Ankh.md gives each project folder its own Hermes identity, config, skills, sessions, and memory while preserving the default global agent elsewhere. |
+| [Competing Autonomous AI Companies](usecases/competing-autonomous-ai-companies.md) | Run two Paperclip companies with Hermes workers against identical tasks, then compare execution evidence and transfer learned skills across teams. |
 
 ## Memory & Context
 
 | Name | Description |
 | --- | --- |
 | [YantrikDB Cognitive Memory (via MCP)](usecases/yantrikdb-memory-mcp.md) | Replace or augment native memory with a dedicated cognitive DB — vector + knowledge graph + temporal decay + contradiction detection. Integrates over MCP, so no Hermes plugin needed. |
 | [Obsidian Vault as a Second Brain](usecases/obsidian-second-brain.md) | Wire Hermes into an Obsidian vault for persistent knowledge — read notes, create daily dashboards, cross-reference tags, maintain a self-improving personal knowledge base. |
+| [Hermes Memory OS](usecases/hermes-memory-os.md) | A seven-layer local memory stack with session search, trusted facts, hybrid vector retrieval, automatic extraction, a self-curating wiki, and ground-truth rules. |
+| [Quantified-Self Life OS](usecases/quantified-self-life-os.md) | Track personal routines and wellness signals, detect evidence-bounded correlations, and deliver scheduled morning, evening, and weekly reflections. |
 
 ## Observability & Operations
 
 | Name | Description |
 | --- | --- |
 | [Hermes Labyrinth Observability](usecases/hermes-labyrinth-observability.md) | A read-only dashboard plugin that turns prompts, tool calls, failures, approvals, model switches, cron runs, and memory hits into auditable agent journeys. |
 | [Deterministic Record and Replay of Hermes Runs](usecases/deterministic-record-and-replay.md) | Records a run at the process and socket boundary, replays it offline byte for byte with the network off, and forks it from any checkpoint onto a different model. Needs no plugin and no change to `config.yaml`. |
+| [Autonomous SRE Incident Commander](usecases/autonomous-sre-incident-commander.md) | Detect, triage, diagnose, safely remediate, verify, document, and learn from Linux, Docker, or Kubernetes incidents. |
+| [Local API Trace Debugging](usecases/local-api-trace-debugging.md) | Run Hermes through claude-tap to inspect provider requests, prompts, tools, streaming responses, token usage, and request diffs in a local viewer. |
 
 ## Security & Assurance
 
 | Name | Description |
 | --- | --- |
 | [Autonomous Penetration Testing with Kali](usecases/autonomous-penetration-testing.md) | A Kali-backed Hermes skill for authorized security assessments with scope confirmation, playbooks, zero-findings fallback, approval gates, and structured reports. |
+| [Sandboxed Hermes with NVIDIA OpenShell](usecases/hermes-openshell-sandbox.md) | Run Hermes behind out-of-process network, filesystem, syscall, and inference policies while retaining approved memory, skills, cron, and gateway capabilities. |
 
 ## Finance & Trading
 
@@ -160,6 +170,8 @@ Then pick a use case below. The [`/demos`](demos/) folder has three runnable sta
 | [China-Region LLM Endpoints](usecases/china-region-llm-endpoints.md) | Dedicated China-facing provider IDs for Kimi/Moonshot, MiniMax, and z.ai/GLM — auto-probing endpoints, no hand-wired base URLs. Pair with Feishu/WeCom for a fully domestic stack. |
 | [Nous Portal Gateway](usecases/nous-portal-gateway.md) | The maintainers' recommended default — one OAuth login unlocks 300+ models plus a Tool Gateway (web search, image gen, TTS, browser) and Nous Chat. Scoped JWTs, no per-provider key juggling. |
 | [Coding-Tool Subscriptions as Backends (Copilot / Codex / Claude Max)](usecases/oauth-cloud-providers.md) | Reuse existing coding-tool subscriptions as Hermes model backends — Copilot direct + ACP, Codex device-code OAuth, Claude Max extra-usage OAuth — with native auth and 401 one-shot recovery. |
+| [Integrated Self-Hosted Hermes Stack](usecases/integrated-self-hosted-hermes-stack.md) | Bootstrap Hermes with model routing, local inference, search, vector memory, automation, notifications, observability, and monitoring through a tiered Docker stack. |
+| [Persistent Hermes on Hugging Face Spaces](usecases/hermes-on-hugging-face-spaces.md) | Deploy Hermes as a Space with a web dashboard and messaging gateway, synchronizing durable agent state to a private Hugging Face Dataset. |
 
 ## Research & Training
 
 
@@ -32,13 +32,9 @@ The entries below are derived from the initial fact-check of candidate sources.
 | **Raspberry Pi / edge deployment** | Pattern is reasonable but submission was framework-agnostic. Hermes officially supports Android/Termux on ARM. | A first-person write-up of Hermes on a Pi (or Termux) with specific model, config, and operational notes. |
 | **Sim (visual workflow engine) × Hermes** | Sim is a real project but the submission didn't document the Hermes integration path. | A write-up showing Sim's MCP export consumed by Hermes as an MCP server, with config and a concrete use case. |
 | **Google Workspace Orchestration** | Public Hermes-compatible repo exists, but `BruceLanLan/hermes-google-workspace` is currently below the catalog's GitHub star threshold. | Promote once the repo clears the 10-star gate, or submit under a PR explicitly labeled `nascent` with current maintainer context. |
-| **Nextcloud Cloud Workspace Integration** | Public Hermes skill repo exists, but `adnw-vinc/hermes-nextcloud` is currently below the catalog's GitHub star threshold. | Promote once the repo clears the 10-star gate, or submit under a PR explicitly labeled `nascent` with current maintainer context. |
 | **MCP backend for Claude Desktop / Cursor** | Surfaced in the `mudrii/hermes-agent-docs` community mirror only; not present in the official `programmatic-integration.md` (verified — that doc covers ACP, TUI-gateway JSON-RPC, and the OpenAI-compat API server only). | A link to official Hermes docs/release notes describing an MCP-server mode exposed to external clients (Claude Desktop, Cursor). Distinct from Hermes *consuming* MCP, which is already documented. |
 | **WhyBuddy (product rehearsal engine)** | Real repo, `xiaojilele-glitch/WhyBuddy`, 349★ — description names a "universal product spec & rehearsal engine" but does not name Hermes in the short description. | A README read confirming Hermes is the substrate (gateway/skills/delegation), then a usecase file. Star gate already cleared. |
-| **hermes-life-os (self-evolving personal OS)** | Real repo, `Lethe044/hermes-life-os`, 129★ — description explicitly names "Memory + Cron + Atropos RL", a direct Hermes-primitive descendant. Strong candidate. | A README read to capture the actual workflow, then a usecase file. Star gate cleared; likely promotes on first read. |
 | **Aether-desktop-orchestrator (multi-AI desktop automation)** | Real repo, `pnganga813-commits/Aether-desktop-orchestrator`, 74★ — description tags it "Hermes Agent 2026: Multi-AI Desktop Automation Platform" using computer-use. | A README read confirming the Hermes integration path (computer-use skill usage), then a usecase file. Star gate cleared. |
-| **Ankh.md (multi-agent swarm framework)** | Real repo, `Abruptive/Ankh.md`, 69★ — description states "Ankh.md is a Hermes Agent (Nous Research) launch exclusive." | A README read to extract the swarm pattern + Hermes primitives used, then a usecase file. Star gate cleared. |
-| **gladiator (two-AI-company GitHub-star battle)** | Real repo, `runtimenoteslabs/gladiator`, 58★ — description: "Two zero-human AI companies battle for GitHub stars using Hermes Agent + Paperclip. Nous Research Hackathon 2026." | A README read to capture the Hermes + Paperclip orchestration shape, then a usecase file. Star gate cleared. |
 
 See the triage files for full rationale:
 - [`submissions-triage-2026-04.md`](./submissions-triage-2026-04.md) — first batch (20 candidates)
 
@@ -0,0 +1,70 @@
+# Autonomous SRE Incident Commander
+
+**Class:** Independent deployment · **Confidence:** High · **Demo status:** Runnable demo + tested watchdog
+
+## Pain Point
+
+During an infrastructure incident, an on-call engineer must collect signals, determine severity, correlate logs, choose a safe response, confirm recovery, and write a post-incident report. Delays and incomplete handoffs increase mean time to recovery.
+
+## What It Does
+
+Hermes Incident Commander implements a `DETECT → TRIAGE → DIAGNOSE → REMEDIATE → VERIFY → DOCUMENT → LEARN` loop for Linux, Docker, and Kubernetes environments. It uses Hermes memory for topology and incident history, cron for health checks, gateway channels for alerts, subagents for parallel service investigation, and skill creation for prevention playbooks.
+
+The repository also includes a standalone watchdog, adaptive time-of-day thresholds, local SQLite/FTS incident search, Prometheus metrics, PagerDuty resolution, dry-run mode, and allowlisted remediation.
+
+## Setup
+
+Clone the repository, install Hermes, and copy the skill:
+
+```bash
+git clone https://github.com/Lethe044/hermes-incident-commander.git
+cd hermes-incident-commander
+
+curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
+hermes setup
+hermes gateway setup
+
+cp -r skills/incident-commander ~/.hermes/skills/
+```
+
+Run the included simulated incident before pointing anything at a real host:
+
+```bash
+pip install anthropic rich
+python demo/demo_incident.py --scenario disk-full-logs
+```
+
+For the standalone watchdog, preview every proposed action first:
+
+```bash
+pip install -e .
+python -m monitor.watchdog --dry-run
+```
+
+## Prompts
+
+The repository documents this Hermes scheduling prompt:
+
+```text
+Set up incident monitoring: run a health check every 5 minutes and alert me
+on Telegram if anything is P0 or P1. Send me a daily briefing at 08:00.
+```
+
+## Skills Needed
+
+- `incident-commander` skill
+- Terminal access to the monitored Linux, Docker, or Kubernetes environment
+- Hermes cron and a configured alert gateway
+- Optional PagerDuty, Prometheus/Grafana, or cloud-provider MCP integrations
+
+## Notes
+
+- Start in the included demo or `--dry-run` mode. Real infrastructure access changes this from an educational workflow into a high-impact operator.
+- The standalone watchdog's auto-remediation is opt-in and allowlisted. Preserve that boundary instead of granting unrestricted shell repair.
+- The repository distinguishes safe actions, warned actions, and destructive actions that require explicit approval.
+
+## Sources
+
+- Incident Commander repository: <https://github.com/Lethe044/hermes-incident-commander>
+- Hermes Agent repository: <https://github.com/NousResearch/hermes-agent>
+
@@ -0,0 +1,57 @@
+# Competing Autonomous AI Companies
+
+**Class:** Independent deployment · **Confidence:** Medium-High · **Demo status:** Runnable hackathon demonstration
+
+## Pain Point
+
+Multi-agent demonstrations often show activity but provide little evidence about whether agents completed work, learned reusable techniques, or benefited from another team's discoveries. It is difficult to compare organizational strategies under the same task and starting conditions.
+
+## What It Does
+
+Gladiator creates two autonomous Paperclip companies with different strategies and identical starter repositories. Nine Hermes workers execute ten tasks while a dashboard records task progress, Git commits, skill versions, memory growth, session continuity, and an audit trail.
+
+After the competition, a merge phase transfers skills between the teams and records cross-agent use. This makes the project useful as an orchestration and learning demonstration, not as evidence that its synthetic GitHub-star score predicts real product adoption.
+
+## Setup
+
+The full demo requires Python 3.11+, Node.js 20+, pnpm 9+, PostgreSQL 16+, Hermes, Paperclip, and an Anthropic API key. After configuring the Hermes Paperclip adapter:
+
+```bash
+git clone https://github.com/runtimenoteslabs/gladiator.git
+cd gladiator
+cp .env.example .env
+# add the documented credentials and Paperclip connection values
+```
+
+Start Paperclip, the dashboard, and its watcher using the repository's setup guide, then open `/landing` and select **LAUNCH DEMO**. The demo initializes fresh repositories, wakes the workers, runs the timed competition, and enables the post-result skill-transfer merge.
+
+## Prompts
+
+Gladiator's work prompts are predefined Paperclip issues rather than a single interactive prompt. The source documents the adapter invoking Hermes as:
+
+```text
+hermes chat -q "<Paperclip task prompt>" -Q
+```
+
+Use the supplied task set so both companies receive the same work and the comparison remains meaningful.
+
+## Skills Needed
+
+- Hermes Agent and `hermes-paperclip-adapter`
+- Paperclip with PostgreSQL
+- Python, Node.js, and pnpm
+- Anthropic model access for the documented configuration
+- Gladiator dashboard, watcher, and SQLite evidence database
+
+## Notes
+
+- The repository estimates roughly $5–6 in API cost for a complete run. Set provider budgets before launching repeated experiments.
+- The documented adapter version includes local patch notes; check whether newer releases have fixed them before modifying dependencies.
+- Treat the projected-star formula as a game score. Validate software quality with tests and human review before merging generated work.
+
+## Sources
+
+- Gladiator repository: <https://github.com/runtimenoteslabs/gladiator>
+- Hermes Agent repository: <https://github.com/NousResearch/hermes-agent>
+- Paperclip repository: <https://github.com/paperclipai/paperclip>
+
@@ -0,0 +1,60 @@
+# Hermes Memory OS
+
+**Class:** Ecosystem integration · **Confidence:** High · **Demo status:** Runnable local stack
+
+## Pain Point
+
+Long-running Hermes users accumulate decisions, preferences, project history, and facts across many sessions. Flat memory files alone eventually become hard to search and expensive to inject, while retrieved context can still be ignored or redundantly rediscovered by the model.
+
+## What It Does
+
+Memory OS adds seven cooperating layers to Hermes:
+
+1. Workspace memory files injected every turn.
+2. FTS5 search across session history.
+3. Structured facts with entity resolution and trust scoring.
+4. A modified Icarus/Fabric layer for cross-session extraction and recall.
+5. Qdrant hybrid semantic search with lexical and SQLite fallbacks.
+6. A self-curating wiki ingested into the vector store.
+7. A `SOUL.md` and rulebook hierarchy that tells the agent how to treat injected ground truth.
+
+Hooks recall relevant context before model calls, extract learnings afterward, deduplicate per session, and skip trivial social messages.
+
+## Setup
+
+Requirements are Hermes Agent, Docker, and Python 3.11+. The repository provides a one-command installer:
+
+```bash
+curl -sSL https://raw.githubusercontent.com/ClaudioDrews/memory-os/main/setup.sh | bash
+```
+
+The installer brings up Qdrant, Redis, and the ARQ worker, initializes the SQLite stores, and installs the Hermes-side components. The repository also provides a manual installation guide and smoke tests for troubleshooting.
+
+## Prompts
+
+Memory OS operates through Hermes hooks; it does not need a special prompt for every recall. A useful verification pattern is to record a concrete project decision in one session, start a new session, and ask:
+
+```text
+What decision did we make about this project's deployment, and why?
+```
+
+The expected behavior is retrieval from existing memory, not re-investigation with tools.
+
+## Skills Needed
+
+- Hermes Agent hook/plugin support
+- Docker and Python 3.11+
+- Qdrant, Redis, ARQ worker, and the bundled SQLite databases
+- Optional local LLM provider; the memory infrastructure is provider-agnostic
+
+## Notes
+
+- This is substantially heavier than native `MEMORY.md`, `USER.md`, and session search. Use it when cross-session scale and structured recall justify additional services.
+- The project is local-first, but the selected LLM provider may still receive recalled context. Local memory storage does not automatically make inference local.
+- Review what the automatic extractor stores; durable memory can preserve sensitive or obsolete information.
+
+## Sources
+
+- Memory OS repository: <https://github.com/ClaudioDrews/memory-os>
+- Hermes Agent repository: <https://github.com/NousResearch/hermes-agent>
+