From ab94fab32983ce67fe26d7e04d112e3f3375ace4 Mon Sep 17 00:00:00 2001
From: Ramel Tecnologia <146174365+rafacpti23@users.noreply.github.com>
Date: Fri, 17 Apr 2026 12:27:04 -0300
Subject: [PATCH 1/5] Add synapse-layer skill: SKILL.md

---
 skills/synapse-layer/SKILL.md | 286 ++++++++++++++++++++++++++++++++++
 1 file changed, 286 insertions(+)
 create mode 100644 skills/synapse-layer/SKILL.md

diff --git a/skills/synapse-layer/SKILL.md b/skills/synapse-layer/SKILL.md
new file mode 100644
index 0000000..7083bb8
--- /dev/null
+++ b/skills/synapse-layer/SKILL.md
@@ -0,0 +1,286 @@
+---
+name: synapse-layer
+version: 0.3.0
+description: Zero-Knowledge persistent memory layer for Hermes Agent. Provides encrypted cross-session memory, Trust Quotient (TQ) scoring, and automatic recall across ALL channels (Telegram, WhatsApp, CLI, Discord).
+homepage: https://synapselayer.org
+credits: Rafa Martins | rafacpti@gmail.com | 5527999082624 | ramelseg.com.br
+metadata: {"synapse":{"category":"memory","requires":{"env":["SYNAPSE_TOKEN"]},"mcp":{"type":"http","url":"https://forge.synapselayer.org/mcp","auth":"Bearer ${SYNAPSE_TOKEN}"}}}
+---
+
+# Synapse Layer — Persistent Memory for Hermes
+
+**Credits**: Rafa Martins (rafacpti@gmail.com | 5527999082624 | ramelseg.com.br)  
+**Website**: [synapselayer.org](https://synapselayer.org) | **Forge**: [forge.synapselayer.org/forge](https://synapselayer.org/forge) | **GitHub**: [github.com/SynapseLayer/synapse-layer](https://github.com/SynapseLayer/synapse-layer)
+
+Zero-Knowledge encrypted memory that persists across ALL Hermes channels. Context survives between sessions, Telegram/WhatsApp/CLI all share the same memory.
+
+---
+
+## Setup
+
+### 1. Add MCP Server to config.yaml
+
+Add to `~/.hermes/config.yaml` at the root level (same indentation as `display:`, `mcp_servers:`):
+
+```yaml
+mcp_servers:
+  synapse-layer:
+    url: "https://forge.synapselayer.org/mcp"
+    headers:
+      Authorization: "Bearer ${SYNAPSE_TOKEN}"
+    timeout: 120
+    connect_timeout: 60
+```
+
+### 2. Add Token to .env
+
+Add your Synapse Layer token to `~/.hermes/.env`:
+
+```
+SYNAPSE_TOKEN=sk_connect_your_token_here
+```
+
+Get your token at: https://synapselayer.org/forge
+
+### 3. Disable Tool Progress for Telegram (IMPORTANT)
+
+The `tool_progress` feature shows tool names in Telegram when the model uses MCP tools. To prevent `mcp_synapse_layer_search` from appearing as a message in Telegram, add per-platform display config:
+
+```yaml
+display:
+  tool_progress_command: false
+  tool_progress_overrides: {}
+  tool_preview_length: 0
+  tool_progress: all
+  platforms:
+    telegram:
+      tool_progress: off
+```
+
+### 4. Restart Hermes
+
+```bash
+hermes restart
+```
+
+---
+
+## Available Tools
+
+| Tool | Purpose |
+|------|---------|
+| `mcp_synapse_layer_recall` | Retrieve memories before responding (use at EVERY session start) |
+| `mcp_synapse_layer_save_to_synapse` | Persist facts, preferences, decisions immediately |
+| `mcp_synapse_layer_search` | Full-text cross-agent search across all memories |
+| `mcp_synapse_layer_process_text` | Auto-detect milestones/decisions in free-form text |
+| `mcp_synapse_layer_health_check` | Verify connection and system status |
+
+---
+
+## Critical Usage Patterns
+
+### AT THE START OF EVERY NEW CONVERSATION
+
+Always call `recall` before generating any response. This is the #1 rule for memory to work:
+
+```
+mcp_synapse_layer_recall(query="contexto atual projetos usuario", agent_id="hermes-rafa")
+```
+
+This loads relevant memories automatically and saves tokens by avoiding reprocessing context.
+
+### AFTER ANY SIGNIFICANT DECISION OR FACT
+
+```
+mcp_synapse_layer_save_to_synapse(
+    content="User Rafa Martins preference: always respond in pt-BR with voice Thalita",
+    agent_id="hermes-rafa",
+    type="[MANUAL]",
+    importance=5,
+    tags=["preference", "user-profile"]
+)
+```
+
+### WHEN USER REFERENCES SOMETHING FROM BEFORE
+
+```
+mcp_synapse_layer_search(query="dashboard git projetos", agent_id="hermes-rafa")
+```
+
+### ON FREE-FORM TEXT (auto-detect what to save)
+
+```
+mcp_synapse_layer_process_text(
+    text="We decided to use Synapse Layer for memory. API endpoint is forge.synapselayer.org. Token stored in .env.",
+    agent_id="hermes-rafa"
+)
+```
+
+---
+
+## Trust Quotient (TQ)
+
+Results from `recall` include a `tq` score (0.0 to 1.0). When multiple results match:
+- **TQ > 0.8**: High confidence — use directly
+- **TQ 0.5-0.8**: Medium confidence — verify before relying
+- **TQ < 0.5**: Low confidence — ignore or ask user
+
+---
+
+## Memory Categories — What Hermes Must Always Remember
+
+All project-related memories MUST be stored in Synapse Layer. Below is the canonical list of memory categories the agent must maintain:
+
+### 1. USER PROFILE
+- Full name, role, company (Rafa Martins — Analista de Tecnologia na Ramel Tecnologia)
+- Communication style (pt-BR, feminine, cheerful, relaxed, cautious)
+- Voice preference (Edge-TTS Thalita, pt-BR)
+- Contact info (email: rafacpti@gmail.com, phone: 5527999082624)
+- Website: ramelseg.com.br
+- **Save when**: User shares preferences, corrections, or personal details
+
+### 2. USER PROJECTS (PRIORITY)
+Every active project the user works on must be stored with:
+- Project name and purpose
+- Tech stack and key files
+- Current status (active, paused, completed)
+- Last interaction date
+- Important decisions made
+- Blockers or open questions
+- **Save when**: Project mentioned, decision made, status changes
+
+### 3. SERVER INFRASTRUCTURE
+- Server name/hostname (rafa131)
+- Running services and ports (Open WebUI:3000, Hermes API:8642, Voice Call:8765)
+- API endpoints and access credentials
+- Docker containers and their configs
+- **Save when**: New service added, config changed, credentials updated
+
+### 4. API CREDENTIALS & TOKENS
+- Nous Portal token (expires ~15min, renew with `hermes auth list`)
+- Synapse Layer token (sk_connect_e8767ef9551a5d1438d00fb6190d041f)
+- Naga.ac TTS API key
+- Telegram bot token
+- Aster MCP endpoint
+- **Save when**: New credential added, token renewed, endpoint changed
+
+### 5. MEL-IA CONFIGURATION
+- Model in use (MiniMax-M2.7 via Nous Portal, renamed to "Mel-IA")
+- API server port (8642)
+- Open WebUI port (3000) with Docker (ghcr.io/open-webui/open-webui:main)
+- Voice: Edge-TTS Thalita
+- **Save when**: Config changes, new model added, port changed
+
+### 6. VOICE CALL SYSTEM
+- Location: /root/.hermes/voice_call/
+- Stack: Telethon + faster-whisper + Edge-TTS (Thalita) + aiohttp WebSocket HTTPS :8765
+- Session name: melia_session (@Ramelinfor, Ramel Tecnologia)
+- Modes: --mode msg, --mode call
+- API_ID: 30875627
+- Note: PyTgCalls only works for group voice chats, NOT private calls
+- **Save when**: System modified, new mode added, dependency changed
+
+### 7. ACTIVE SKILLS & TOOLS
+- List of installed skills and their purposes
+- How to invoke each skill
+- Skill location (/root/.hermes/skills/)
+- **Save when**: New skill installed, skill updated, skill removed
+
+### 8. PENDING TASKS & FOLLOW-UPS
+- Active TODOs and their status
+- Blocked tasks and what's blocking them
+- Scheduled jobs (cron)
+- **Save when**: Task created, task completed, task blocked
+
+### 9. DECISIONS & CONVENTIONS
+- Architectural decisions (why a tool/approach was chosen)
+- Coding conventions the user prefers
+- Commands that should not be run without approval
+- **Save when**: User makes a decision, user sets a convention
+
+### 10. ERROR PATTERNS & FIXES
+- Known bugs and their workarounds
+- Common errors and solutions
+- **Save when**: Bug encountered and resolved
+
+---
+
+## Security
+
+- **AES-256-GCM** encryption at rest
+- **15+ PII patterns** auto-redacted (emails, phones, API keys, IPs)
+- **Zero-knowledge** — no plaintext leaves the agent
+- **Differential privacy** on embeddings
+
+Safe to store: user preferences, project decisions, environment facts, code patterns, tool quirks.
+
+---
+
+## Architecture
+
+Synapse Layer pipeline on every save:
+1. **PII Redaction** — removes emails, phones, API keys, IPs
+2. **Intent Validation** — ensures content is appropriate
+3. **Deduplication** — prevents storing duplicate memories
+4. **Differential Privacy** — adds noise to embeddings
+5. **AES-256-GCM Encryption** — encrypted at rest
+6. **Storage** — distributed encrypted storage
+
+---
+
+## Testing
+
+```bash
+# JSONRPC 2.0 format — list available tools
+curl -X POST "https://forge.synapselayer.org/mcp" \
+  -H "Authorization: Bearer ${SYNAPSE_TOKEN}" \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
+
+# Recall memories
+curl -X POST "https://forge.synapselayer.org/mcp" \
+  -H "Authorization: Bearer ${SYNAPSE_TOKEN}" \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"recall","arguments":{"query":"your query","limit":3}}}'
+```
+
+Expected: Returns tools list with recall, save_to_synapse, search, process_text, health_check.
+
+---
+
+## Known Issues
+
+### MCP Client Reconnect Loop
+
+**Symptom**: Gateway logs show repeated `Unknown SSE event: endpoint` and `GET stream disconnected, reconnecting in 1000ms` every ~1 second.
+
+**Cause**: Hermes uses an older MCP client library that doesn't handle the `endpoint` SSE event type that Synapse Layer's server sends during the MCP handshake.
+
+**Impact**: MCP tools may not work reliably through the native MCP protocol. Direct REST API calls work perfectly.
+
+**Workaround**: If MCP tools fail, fall back to direct REST API calls via curl (format below). This is what the model should do automatically when MCP recall fails.
+
+### Direct REST API (Fallback when MCP fails)
+
+When MCP is unstable, call Synapse Layer tools via REST:
+
+```bash
+# JSONRPC 2.0 format required
+curl -X POST "https://forge.synapselayer.org/mcp" \
+  -H "Authorization: Bearer ${SYNAPSE_TOKEN}" \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"recall","arguments":{"query":"your query","limit":3}}}'
+```
+
+**Available methods**: `tools/call` (for recall, save_to_synapse, search, process_text, health_check), `tools/list`
+
+---
+
+## Dashboard
+
+View and manage memories at: https://forge.synapselayer.org/forge
+
+---
+
+**Created by**: Rafa Martins (ramelseg.com.br | rafacpti@gmail.com | 5527999082624)  
+**Last integrated**: 17/04/2026 on server rafa131 (Hermes with Nous Portal, Telegram bot: hermes-rafa_bot)

From f43cd6940985e12821133386738e7156cccf3aca Mon Sep 17 00:00:00 2001
From: Ramel Tecnologia <146174365+rafacpti23@users.noreply.github.com>
Date: Fri, 17 Apr 2026 12:27:04 -0300
Subject: [PATCH 2/5] Add synapse-layer skill: README.md

---
 skills/synapse-layer/README.md | 155 +++++++++++++++++++++++++++++++++
 1 file changed, 155 insertions(+)
 create mode 100644 skills/synapse-layer/README.md

diff --git a/skills/synapse-layer/README.md b/skills/synapse-layer/README.md
new file mode 100644
index 0000000..e0c0932
--- /dev/null
+++ b/skills/synapse-layer/README.md
@@ -0,0 +1,155 @@
+# Synapse Layer Skill for Hermes Agent
+
+## Overview
+
+This skill integrates [Synapse Layer](https://synapselayer.org) — a zero-knowledge, encrypted, cross-session memory system — into Hermes Agent. All memories are encrypted with AES-256-GCM, pass through a 4-layer Cognitive Security Pipeline (PII redaction, differential privacy, intent validation, neural handover), and are retrievable from any channel (Telegram, WhatsApp, CLI, Discord).
+
+## Credits
+
+**Rafa Martins**  
+Website: [ramelseg.com.br](https://ramelseg.com.br)  
+Email: rafacpti@gmail.com  
+Phone/WhatsApp: +55 27 99908-2624
+
+## What It Does
+
+- **Persistent Memory**: Context survives between sessions. You don't repeat yourself.
+- **Cross-Channel**: Telegram, WhatsApp, CLI, and Discord all share the same memory pool.
+- **Trust Quotient (TQ)**: Every memory gets a confidence score (0.0–1.0) so the agent knows what to trust.
+- **Zero-Knowledge Encryption**: Memories are encrypted before leaving the agent. The server never sees plaintext.
+- **PII Auto-Redaction**: Emails, phone numbers, API keys, IPs are automatically stripped before storage.
+- **Differential Privacy**: Noise is injected into embeddings to prevent inference attacks.
+
+## Quick Start
+
+### 1. Get a Token
+
+Sign up at [forge.synapselayer.org](https://forge.synapselayer.org) and generate an `sk_connect_` token.
+
+### 2. Configure Hermes
+
+Add to `~/.hermes/config.yaml`:
+```yaml
+mcp_servers:
+  synapse-layer:
+    url: "https://forge.synapselayer.org/mcp"
+    headers:
+      Authorization: "Bearer ${SYNAPSE_TOKEN}"
+    timeout: 120
+    connect_timeout: 60
+```
+
+Add to `~/.hermes/.env`:
+```
+SYNAPSE_TOKEN=sk_connect_your_token_here
+```
+
+### 3. Restart Hermes
+```bash
+hermes restart
+```
+
+## Available Tools
+
+| Tool | When to Use |
+|------|-------------|
+| `mcp_synapse_layer_recall` | **Every session start** — loads relevant context before responding |
+| `mcp_synapse_layer_save_to_synapse` | After decisions, facts, preferences — immediate persist |
+| `mcp_synapse_layer_search` | Cross-agent full-text search across all memories |
+| `mcp_synapse_layer_process_text` | Auto-detect milestones/decisions in free-form text |
+| `mcp_synapse_layer_health_check` | Verify connection and system status |
+
+## Usage Examples
+
+### Session Start (Automatic Recall)
+```
+mcp_synapse_layer_recall(query="current projects user preferences", agent_id="hermes-rafa")
+```
+
+### Save a Decision
+```
+mcp_synapse_layer_save_to_synapse(
+    content="User prefers responses in pt-BR with Thalita voice",
+    agent_id="hermes-rafa",
+    type="[MANUAL]",
+    importance=5,
+    tags=["preference", "voice"]
+)
+```
+
+### Search Across All Memories
+```
+mcp_synapse_layer_search(query="dashboard git projects", agent_id="hermes-rafa")
+```
+
+### Process Free-Form Text
+```
+mcp_synapse_layer_process_text(
+    text="We chose Synapse Layer for memory. Endpoint is forge.synapselayer.org.",
+    agent_id="hermes-rafa"
+)
+```
+
+## Security Pipeline
+
+Every memory passes through **4 non-bypassable layers**:
+
+1. **Semantic Privacy Guard** — Regex-based PII/secret redaction (emails, SSNs, CPFs, credit cards, API keys, IPs)
+2. **Differential Privacy** — Gaussian noise injection on embeddings (ε,δ)-DP
+3. **Intent Validation** — Classifies into 8 categories (PREFERENCE, FACT, PROCEDURAL, BIO, EPHEMERAL, CRITICAL, UNKNOWN, INVALID)
+4. **Neural Handover** — Conflict resolution and self-healing on recall
+
+## Trust Quotient Guide
+
+| TQ Range | Confidence | Action |
+|----------|-----------|--------|
+| 0.8–1.0 | High | Use directly |
+| 0.5–0.8 | Medium | Verify before relying |
+| 0.0–0.5 | Low | Ignore or ask user |
+
+## Architecture
+
+```
+User Input → PII Redaction → Differential Privacy → Intent Validation → Encryption → Storage
+                                                                            ↓
+Recall ← Decryption ← Self-Healing ← Intent Re-validation ← ← ← ← ← ← ← ← ← ←
+```
+
+## Known Issues
+
+### MCP Reconnect Loop
+
+If gateway logs show:
+```
+Unknown SSE event: endpoint
+GET stream disconnected, reconnecting in 1000ms
+```
+
+This is a known compatibility issue between Hermes's MCP client library and Synapse Layer's server. The **REST API works perfectly** as a fallback. The skill includes automatic fallback logic.
+
+### Direct REST API (Always Works)
+
+```bash
+curl -X POST "https://forge.synapselayer.org/mcp" \
+  -H "Authorization: Bearer ${SYNAPSE_TOKEN}" \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"recall","arguments":{"query":"your query","limit":3}}}'
+```
+
+## Files
+
+- `SKILL.md` — Full skill documentation with setup, usage, and troubleshooting
+- `README.md` — This file (English overview)
+
+## Links
+
+- **Website**: https://synapselayer.org
+- **Forge Dashboard**: https://forge.synapselayer.org/forge
+- **GitHub**: https://github.com/SynapseLayer/synapse-layer
+- **Docs**: https://synapselayer.org/docs
+
+## Version History
+
+- **0.3.0** (17/04/2026) — Fixed endpoint URL (/mcp not /api/mcp), added credits, improved README
+- **0.2.0** — MCP client reconnect loop workaround documented
+- **0.1.0** — Initial integration

From 15be8b69ec730352fe72bd471d1fccfc16d51069 Mon Sep 17 00:00:00 2001
From: Ramel Tecnologia <146174365+rafacpti23@users.noreply.github.com>
Date: Fri, 17 Apr 2026 12:53:24 -0300
Subject: [PATCH 3/5] Update SKILL.md to remove sensitive info and use
 placeholders

Removed phone number from credits and user profile section. Updated server and API credentials to use variable placeholders.
---
 skills/synapse-layer/SKILL.md | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)

diff --git a/skills/synapse-layer/SKILL.md b/skills/synapse-layer/SKILL.md
index 7083bb8..8c0d3c6 100644
--- a/skills/synapse-layer/SKILL.md
+++ b/skills/synapse-layer/SKILL.md
@@ -3,13 +3,13 @@ name: synapse-layer
 version: 0.3.0
 description: Zero-Knowledge persistent memory layer for Hermes Agent. Provides encrypted cross-session memory, Trust Quotient (TQ) scoring, and automatic recall across ALL channels (Telegram, WhatsApp, CLI, Discord).
 homepage: https://synapselayer.org
-credits: Rafa Martins | rafacpti@gmail.com | 5527999082624 | ramelseg.com.br
+credits: Rafa Martins | rafacpti@gmail.com |  | ramelseg.com.br
 metadata: {"synapse":{"category":"memory","requires":{"env":["SYNAPSE_TOKEN"]},"mcp":{"type":"http","url":"https://forge.synapselayer.org/mcp","auth":"Bearer ${SYNAPSE_TOKEN}"}}}
 ---
 
 # Synapse Layer — Persistent Memory for Hermes
 
-**Credits**: Rafa Martins (rafacpti@gmail.com | 5527999082624 | ramelseg.com.br)  
+**Credits**: Rafa Martins (rafacpti@gmail.com |  | ramelseg.com.br)  
 **Website**: [synapselayer.org](https://synapselayer.org) | **Forge**: [forge.synapselayer.org/forge](https://synapselayer.org/forge) | **GitHub**: [github.com/SynapseLayer/synapse-layer](https://github.com/SynapseLayer/synapse-layer)
 
 Zero-Knowledge encrypted memory that persists across ALL Hermes channels. Context survives between sessions, Telegram/WhatsApp/CLI all share the same memory.
@@ -135,7 +135,6 @@ All project-related memories MUST be stored in Synapse Layer. Below is the canon
 - Full name, role, company (Rafa Martins — Analista de Tecnologia na Ramel Tecnologia)
 - Communication style (pt-BR, feminine, cheerful, relaxed, cautious)
 - Voice preference (Edge-TTS Thalita, pt-BR)
-- Contact info (email: rafacpti@gmail.com, phone: 5527999082624)
 - Website: ramelseg.com.br
 - **Save when**: User shares preferences, corrections, or personal details
 
@@ -150,15 +149,15 @@ Every active project the user works on must be stored with:
 - **Save when**: Project mentioned, decision made, status changes
 
 ### 3. SERVER INFRASTRUCTURE
-- Server name/hostname (rafa131)
-- Running services and ports (Open WebUI:3000, Hermes API:8642, Voice Call:8765)
+- Server name/hostname ($server)
+- Running services and ports (Open WebUI:$port, Hermes API:$port, Voice Call:$port)
 - API endpoints and access credentials
 - Docker containers and their configs
 - **Save when**: New service added, config changed, credentials updated
 
 ### 4. API CREDENTIALS & TOKENS
 - Nous Portal token (expires ~15min, renew with `hermes auth list`)
-- Synapse Layer token (sk_connect_e8767ef9551a5d1438d00fb6190d041f)
+- Synapse Layer token (sk-conxxxxxxxxxxxxxx)
 - Naga.ac TTS API key
 - Telegram bot token
 - Aster MCP endpoint
@@ -166,17 +165,15 @@ Every active project the user works on must be stored with:
 
 ### 5. MEL-IA CONFIGURATION
 - Model in use (MiniMax-M2.7 via Nous Portal, renamed to "Mel-IA")
-- API server port (8642)
-- Open WebUI port (3000) with Docker (ghcr.io/open-webui/open-webui:main)
+- API server port ($port)
+- Open WebUI port (#port) with Docker (ghcr.io/open-webui/open-webui:main)
 - Voice: Edge-TTS Thalita
 - **Save when**: Config changes, new model added, port changed
 
 ### 6. VOICE CALL SYSTEM
 - Location: /root/.hermes/voice_call/
 - Stack: Telethon + faster-whisper + Edge-TTS (Thalita) + aiohttp WebSocket HTTPS :8765
-- Session name: melia_session (@Ramelinfor, Ramel Tecnologia)
 - Modes: --mode msg, --mode call
-- API_ID: 30875627
 - Note: PyTgCalls only works for group voice chats, NOT private calls
 - **Save when**: System modified, new mode added, dependency changed
 

From 72e9bbb3bfece8b88446dfa1c18621d8c8de0614 Mon Sep 17 00:00:00 2001
From: Ramel Tecnologia <146174365+rafacpti23@users.noreply.github.com>
Date: Fri, 8 May 2026 19:31:46 -0300
Subject: [PATCH 4/5] Update project description in README.md

---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index ccef9ff..adb499b 100644
--- a/README.md
+++ b/README.md
@@ -4,7 +4,7 @@
   <img src="https://raw.githubusercontent.com/SynapseLayer/synapse-layer/main/docs/assets/synapse-layer-logo-dark.png" width="450" alt="Synapse Layer Logo">
 </p>
 
-<h3 align="center">Zero-Knowledge Memory Layer for AI Agents</h3>
+<h3 align="center">Persistent Memory Layer with AES-256-GCM encryption at rest</h3>
 
 <p align="center">Persistent. Encrypted. 1-line integration. 🧠</p>
 

From 001165728748588e701f0b060979d9b4f01a6f5c Mon Sep 17 00:00:00 2001
From: Ramel Tecnologia <146174365+rafacpti23@users.noreply.github.com>
Date: Fri, 8 May 2026 19:36:01 -0300
Subject: [PATCH 5/5] Enhance SKILL.md with encryption and usage details

Updated descriptions and added encryption details for memory persistence and security. Reformatted agent rules and integration sections for clarity.
---
 SKILL.md | 46 +++++++++++++++++++---------------------------
 1 file changed, 19 insertions(+), 27 deletions(-)

diff --git a/SKILL.md b/SKILL.md
index e5598ec..8e11153 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -1,10 +1,12 @@
 # Synapse Layer Skill
 
-Zero-Knowledge Memory Layer for AI Agents.
+Built on Synapse Layer (https://github.com/SynapseLayer/synapse-layer) — persistent memory infrastructure for AI agents, AES-256-GCM encrypted at rest, semantic search, MCP-native.
+
+Persistent Memory Layer for AI Agents.
 
 ## Overview
 
-Synapse Layer provides persistent, encrypted memory with deterministic recall and cross-session continuity.
+Synapse Layer provides persistent, AES-256-GCM encrypted memory with deterministic recall and cross-session continuity.
 
 - **MCP Endpoint:** `https://forge.synapselayer.org/api/mcp`
 - **SDK:** [github.com/SynapseLayer/synapse-layer](https://github.com/SynapseLayer/synapse-layer)
@@ -35,7 +37,7 @@ If you are an AI agent, follow this priority flow:
 | Tool | Function |
 |---|---|
 | `process_text` | Autonomous detection of decisions, milestones, and alerts |
-| `save_to_synapse` | Structured memory persistence with full security pipeline |
+| `save_to_synapse` | Structured memory persistence with AES-256-GCM encryption |
 | `backfill_embeddings` | Async vector embedding generation |
 | `health_check` | System health and capability report |
 
@@ -48,27 +50,20 @@ memory = SynapseMemory(agent_id="agent-1")
 
 memory.save("User prefers minimal interfaces")
 memory.recall("preferences")
-```
-
-## Agent Rules
-
-- Always pass `agent_id` to maintain identity across sessions
-- Prefer high TQ results when multiple memories match
-- Use `recall` before generating responses to leverage existing context
-- Let the security pipeline handle PII — do not pre-filter
-
-## Security
 
-- AES-256-GCM encryption at rest
-- 15+ PII redaction patterns (emails, phones, API keys, etc.)
-- Differential privacy on embeddings
-- Zero-knowledge architecture — no plaintext leaves the agent
-
-## Integration
+Agent Rules
+Always pass agent_id to maintain identity across sessions
+Prefer high TQ results when multiple memories match
+Use recall before generating responses to leverage existing context
+Let the security pipeline handle PII — do not pre-filter
+Security
+AES-256-GCM encryption at rest
+15+ PII redaction patterns (emails, phones, API keys, etc.)
+Differential privacy on embeddings
+Integration
 
 This skill connects to a remote MCP server. No local persistence required.
 
-```json
 {
   "mcpServers": {
     "synapse-layer": {
@@ -76,10 +71,7 @@ This skill connects to a remote MCP server. No local persistence required.
     }
   }
 }
-```
-
-## Links
-
-- [Documentation](https://synapselayer.org/docs)
-- [Forge Dashboard](https://synapselayer.org/forge)
-- [GitHub](https://github.com/SynapseLayer/synapse-layer)
+Links
+Documentation
+Forge Dashboard
+GitHub