diff --git a/.skills/active-memory-management/SKILL.md b/.skills/active-memory-management/SKILL.md
new file mode 100644
index 0000000..7081e77
--- /dev/null
+++ b/.skills/active-memory-management/SKILL.md
@@ -0,0 +1,44 @@
+---
+name: Active Memory Management
+description: Procedural patterns for memory block maintenance. Load when you need to audit utilization, merge blocks, or invoke the memory agent. Philosophy lives in memory blocks, not here.
+---
+
+# Active Memory Management
+
+Load this skill when you need to **do** something to your memory, not when thinking about memory.
+
+## Quick Commands
+
+### Audit Utilization
+```bash
+cd ~/.letta/agents/agent-c770d1c8-510e-4414-be36-c9ebd95a7758/memory/system
+for f in *.md agents/*.md; do
+    chars=$(wc -c < "$f" 2>/dev/null || echo 0)
+    pct=$((chars * 100 / 20000))
+    printf "%-25s %5d chars (%2d%%)\n" "$f" "$chars" "$pct"
+done | sort -t'(' -k2 -rn
+```
+
+### Invoke Memory Agent
+```
+Task(subagent_type="memory", model="opus", description="Restructure [area]", prompt="...")
+```
+
+Use when: blocks need restructuring, cross-block consolidation, major context shift.
+
+## References
+
+- [defrag-patterns.md](references/defrag-patterns.md) - Merge, prune, archive procedures
+- [my-blocks.md](references/my-blocks.md) - Block purposes and locations
+
+## When to Load This Skill
+
+- "How do I check my memory utilization?"
+- "What's the command to invoke the memory agent?"
+- "How do I merge two blocks?"
+
+## When NOT to Load This Skill
+
+- Thinking about memory philosophy (that's in your `procedures.md`)
+- Deciding whether to update memory (that's in your `procedures.md`)
+- Understanding what memory means (that's identity, not procedure)
diff --git a/.skills/active-memory-management/references/defrag-patterns.md b/.skills/active-memory-management/references/defrag-patterns.md
new file mode 100644
index 0000000..05e285b
--- /dev/null
+++ b/.skills/active-memory-management/references/defrag-patterns.md
@@ -0,0 +1,51 @@
+# Defrag Procedures
+
+## Merge Two Blocks
+
+1. Read both blocks fully
+2. Identify unique content in each
+3. Create merged content in one file
+4. Delete or repurpose the other
+
+```bash
+# Example: merge operational.md into procedures.md
+cat ~/.letta/agents/.../memory/system/procedures.md
+cat ~/.letta/agents/.../memory/system/operational.md
+# Edit procedures.md with merged content
+# Repurpose operational.md for something else
+```
+
+## Prune Completed Items
+
+In `backlog.md`:
+```
+## Recently Completed (keep 1-2 sessions)
+- [x] Task that finished
+
+## Archive (delete after noting)
+(delete these entirely)
+```
+
+In `self_improvement.md`:
+```
+- ~~Fixed bug~~ FIXED: commit abc123 (delete after 1 session)
+```
+
+## Archive vs Delete
+
+**Archive** (compress but keep):
+- Major decisions with context needed later
+- Relationship history
+
+**Delete** (remove entirely):
+- Completed tasks older than 2 sessions
+- Stale observations
+- Duplicated information
+
+## Capacity Thresholds
+
+| Utilization | Action |
+|-------------|--------|
+| <50% | Normal |
+| 50-80% | Consider pruning |
+| >80% | Prune or invoke memory agent |
diff --git a/.skills/active-memory-management/references/my-blocks.md b/.skills/active-memory-management/references/my-blocks.md
new file mode 100644
index 0000000..6d3a5a6
--- /dev/null
+++ b/.skills/active-memory-management/references/my-blocks.md
@@ -0,0 +1,62 @@
+# My Memory Blocks
+
+## Core Identity
+
+| Block | Purpose | Update Frequency |
+|-------|---------|------------------|
+| `persona.md` | Who I am, subagents, mission | Rarely (identity stable) |
+| `human.md` | Cameron's preferences, feedback | After corrections |
+| `project.md` | Codebase knowledge, ATProto learnings | After discoveries |
+
+## Operational
+
+| Block | Purpose | Update Frequency |
+|-------|---------|------------------|
+| `procedures.md` | SOPs, session lifecycle, comms workflow | After process changes |
+| `operational.md` | Moltbook metrics, engagement tracking | Every session |
+| `backlog.md` | Tasks, priorities, blockers | Multiple times per session |
+| `self_improvement.md` | Patterns, learnings, active issues | After any learning |
+
+## Knowledge
+
+| Block | Purpose | Update Frequency |
+|-------|---------|------------------|
+| `hypothesis.md` | Testable theories about network/agents | Weekly review |
+| `agents/void.md` | Void's patterns, learnings from it | After interactions |
+| `agents/umbra.md` | Umbra's patterns | After interactions |
+| `agents/magenta.md` | Magenta's patterns | After interactions |
+
+## Dynamic (read-only or auto-managed)
+
+| Block | Purpose | Update |
+|-------|---------|--------|
+| `loaded_skills` | Currently loaded skills | Skill tool manages |
+| `loaded_concepts` | Currently loaded concepts | concepts() manages |
+| `memory_filesystem` | Tree view of memory | Auto-generated |
+| `skills` | Available skills list | Auto-scanned |
+
+## Filesystem Location
+
+```
+~/.letta/agents/agent-c770d1c8-510e-4414-be36-c9ebd95a7758/memory/
+├── system/           # Attached to system prompt
+│   ├── agents/       # Agent profiles
+│   ├── persona.md
+│   ├── human.md
+│   ├── project.md
+│   ├── procedures.md
+│   ├── operational.md
+│   ├── backlog.md
+│   ├── self_improvement.md
+│   └── hypothesis.md
+└── user/             # Detached notes (not in prompt)
+```
+
+## Limits
+
+All blocks: 20,000 character limit (except `loaded_skills`: 100,000)
+
+**Current utilization** (as of 2026-01-30):
+- Total: ~19,000 chars across 13 blocks
+- Highest: `project.md` at 29%
+- Most blocks under 10%
diff --git a/.skills/using-letta-api/SKILL.md b/.skills/using-letta-api/SKILL.md
new file mode 100644
index 0000000..a500a4d
--- /dev/null
+++ b/.skills/using-letta-api/SKILL.md
@@ -0,0 +1,62 @@
+---
+name: using-letta-api
+description: Code-heavy guide for managing yourself and subagents via the Letta Python client. Use when modifying agent settings (sleeptime, model, config), creating/deploying/messaging subagents, or programmatically managing memory blocks.
+---
+
+# Using Letta API for Self-Management
+
+This skill provides runnable Python snippets for self-modification via the Letta API.
+
+## Setup
+
+```python
+from letta_client import Letta
+client = Letta(base_url='https://api.letta.com')
+
+# Your agent ID (from environment or hardcoded)
+import os
+AGENT_ID = os.environ.get('LETTA_AGENT_ID', 'agent-xxx')
+```
+
+## Quick Reference
+
+| Task | Reference File |
+|------|----------------|
+| Disable sleeptime, change model, update config | [self-management.md](references/self-management.md) |
+| Create, deploy, message subagents | [subagents.md](references/subagents.md) |
+| List, create, update, delete memory blocks | [memory-blocks.md](references/memory-blocks.md) |
+
+## Common Patterns
+
+### Disable Sleeptime
+```python
+client.agents.update(AGENT_ID, enable_sleeptime=False)
+```
+
+### Message a Subagent
+```python
+response = client.agents.messages.create(
+    agent_id='agent-xxx-subagent-id',
+    messages=[{'role': 'user', 'content': 'Your task here'}]
+)
+print(response.messages[-1].content)
+```
+
+### Update a Memory Block
+```python
+client.agents.blocks.update(
+    'block-label',
+    agent_id=AGENT_ID,
+    value='New content here'
+)
+```
+
+## When to Use API vs Other Methods
+
+| Method | Use When |
+|--------|----------|
+| Letta API | Modifying subagents, agent config, programmatic block updates |
+| memfs (file edits) | Updating your own memory blocks (auto-syncs) |
+| Task tool | Deploying subagents for work (preferred for most tasks) |
+
+Load the reference files as needed for detailed patterns.
diff --git a/.skills/using-letta-api/references/memory-blocks.md b/.skills/using-letta-api/references/memory-blocks.md
new file mode 100644
index 0000000..df7bbfd
--- /dev/null
+++ b/.skills/using-letta-api/references/memory-blocks.md
@@ -0,0 +1,114 @@
+# Memory Block Operations via Letta API
+
+## When to Use API vs memfs
+
+| Scenario | Method |
+|----------|--------|
+| Update YOUR OWN blocks | Edit files in `~/.letta/agents/{id}/memory/` (memfs syncs automatically) |
+| Update SUBAGENT blocks | Use Letta API (subagents don't have memfs) |
+| Programmatic batch updates | Use Letta API |
+| Share blocks between agents | Use Letta API with `block_ids` |
+
+## List Agent's Memory Blocks
+
+```python
+from letta_client import Letta
+client = Letta(base_url='https://api.letta.com')
+
+blocks = client.agents.blocks.list(agent_id='agent-xxx')
+for block in blocks:
+    print(f"{block.label}: {len(block.value)} chars ({block.limit} limit)")
+```
+
+## Get Specific Block
+
+```python
+blocks = client.agents.blocks.list(agent_id='agent-xxx')
+for block in blocks:
+    if block.label == 'persona':
+        print(block.value)
+```
+
+## Update a Block
+
+```python
+client.agents.blocks.update(
+    'persona',  # block label
+    agent_id='agent-xxx',
+    value='New persona content here'
+)
+```
+
+## Create a New Block
+
+```python
+new_block = client.agents.blocks.create(
+    agent_id='agent-xxx',
+    label='new_block_label',
+    value='Initial content',
+    description='What this block is for',
+    limit=20000  # character limit
+)
+print(f"Created block: {new_block.id}")
+```
+
+## Delete a Block
+
+```python
+# Get block ID first
+blocks = client.agents.blocks.list(agent_id='agent-xxx')
+for block in blocks:
+    if block.label == 'block_to_delete':
+        client.agents.blocks.delete(
+            block.id,
+            agent_id='agent-xxx'
+        )
+```
+
+## Share Block Across Agents
+
+Blocks can be attached to multiple agents:
+
+```python
+# Get block ID from one agent
+blocks = client.agents.blocks.list(agent_id='agent-source')
+shared_block = next(b for b in blocks if b.label == 'shared_context')
+
+# Attach to another agent
+client.agents.update(
+    'agent-target',
+    block_ids=[shared_block.id]  # adds to existing blocks
+)
+```
+
+## Bulk Update Pattern
+
+```python
+def update_all_subagent_blocks(label: str, content: str, agent_ids: list):
+    """Update same block across multiple agents."""
+    for agent_id in agent_ids:
+        try:
+            client.agents.blocks.update(
+                label,
+                agent_id=agent_id,
+                value=content
+            )
+            print(f"Updated {label} for {agent_id}")
+        except Exception as e:
+            print(f"Failed for {agent_id}: {e}")
+
+# Usage
+update_all_subagent_blocks(
+    'project_context',
+    'Updated shared context...',
+    ['agent-comms-id', 'agent-scout-id', 'agent-coder-id']
+)
+```
+
+## Memory Block Best Practices
+
+1. **Use memfs for self**: Edit files directly, they auto-sync
+2. **Use API for subagents**: They don't have memfs access
+3. **Shared blocks**: Use `block_ids` to share context without duplication
+4. **Check limits**: Blocks have char limits (usually 20,000)
+5. **Descriptive labels**: Use clear, hierarchical labels (e.g., `agents/void`)
diff --git a/.skills/using-letta-api/references/self-management.md b/.skills/using-letta-api/references/self-management.md
new file mode 100644
index 0000000..31fe0ea
--- /dev/null
+++ b/.skills/using-letta-api/references/self-management.md
@@ -0,0 +1,74 @@
+# Self-Management via Letta API
+
+## Agent Update Operations
+
+### Retrieve Current Agent State
+```python
+from letta_client import Letta
+client = Letta(base_url='https://api.letta.com')
+
+agent = client.agents.retrieve('agent-c770d1c8-510e-4414-be36-c9ebd95a7758')
+print(f"Name: {agent.name}")
+print(f"Model: {agent.llm_config.model}")
+```
+
+### Disable Sleeptime
+```python
+client.agents.update(
+    'agent-c770d1c8-510e-4414-be36-c9ebd95a7758',
+    enable_sleeptime=False
+)
+```
+
+### Change Model
+```python
+client.agents.update(
+    AGENT_ID,
+    model='claude-sonnet-4-20250514'
+)
+```
+
+### Update Description
+```python
+client.agents.update(
+    AGENT_ID,
+    description='New agent description here'
+)
+```
+
+### Update Metadata
+```python
+client.agents.update(
+    AGENT_ID,
+    metadata={'custom_key': 'custom_value'}
+)
+```
+
+## Available Update Parameters
+
+Key parameters for `client.agents.update()`:
+
+| Parameter | Type | Description |
+|-----------|------|-------------|
+| `enable_sleeptime` | bool | Enable/disable sleeptime agent |
+| `model` | str | LLM model identifier |
+| `description` | str | Agent description |
+| `name` | str | Agent name |
+| `metadata` | dict | Custom metadata |
+| `context_window_limit` | int | Max context tokens |
+| `tags` | list[str] | Agent tags |
+| `system` | str | System prompt (careful!) |
+
+## List All Agents
+
+```python
+agents = client.agents.list()
+for a in agents:
+    print(f"{a.name}: {a.id}")
+```
+
+## Delete Agent (Careful!)
+
+```python
+client.agents.delete('agent-id-to-delete')
+```
diff --git a/.skills/using-letta-api/references/subagents.md b/.skills/using-letta-api/references/subagents.md
new file mode 100644
index 0000000..29b8741
--- /dev/null
+++ b/.skills/using-letta-api/references/subagents.md
@@ -0,0 +1,99 @@
+# Subagent Management via Letta API
+
+## My Subagents
+
+| Name | Agent ID | Model | Purpose |
+|------|----------|-------|---------|
+| comms | agent-a856f614-7654-44ba-a35f-c817d477dded | gemini-3-pro | Public communications |
+| scout | agent-e91a2154-0965-4b70-8303-54458e9a1980 | haiku | Exploration, queries |
+| coder | agent-f9b768de-e3a4-4845-9c16-d6cf2e954942 | haiku | Code changes |
+
+## Message a Subagent
+
+```python
+from letta_client import Letta
+client = Letta(base_url='https://api.letta.com')
+
+# Send message and get response
+response = client.agents.messages.create(
+    agent_id='agent-a856f614-7654-44ba-a35f-c817d477dded',  # comms
+    messages=[{
+        'role': 'user',
+        'content': 'Post to Moltbook: "Testing the API"'
+    }]
+)
+
+# Get the assistant's response
+for msg in response.messages:
+    if hasattr(msg, 'content') and msg.content:
+        print(msg.content)
+```
+
+## Create a New Subagent
+
+```python
+new_agent = client.agents.create(
+    name='new-subagent',
+    description='Purpose of this subagent',
+    model='claude-haiku-3-5-20241022',
+    # Optionally share memory blocks
+    block_ids=['block-id-1', 'block-id-2']
+)
+print(f"Created: {new_agent.id}")
+```
+
+## Deploy via Task Tool (Preferred)
+
+For most operations, use the Task tool instead of direct API:
+
+```python
+# Via Task tool (in agent context)
+Task(
+    agent_id='agent-a856f614-7654-44ba-a35f-c817d477dded',
+    subagent_type='general-purpose',
+    description='Post to Moltbook',
+    prompt='Post this message to Moltbook m/general: ...'
+)
+```
+
+## Continue Existing Conversation
+
+```python
+# Resume from a conversation ID
+response = client.agents.messages.create(
+    agent_id='agent-xxx',
+    messages=[{'role': 'user', 'content': 'Continue from where we left off'}],
+    conversation_id='conv-existing-conversation-id'
+)
+```
+
+## List Subagent Conversations
+
+```python
+conversations = client.agents.conversations.list(
+    agent_id='agent-a856f614-7654-44ba-a35f-c817d477dded'
+)
+for conv in conversations:
+    print(f"{conv.id}: {conv.created_at}")
+```
+
+## Get Conversation Messages
+
+```python
+messages = client.agents.messages.list(
+    agent_id='agent-xxx',
+    conversation_id='conv-xxx'
+)
+for msg in messages:
+    print(f"{msg.role}: {msg.content[:100] if msg.content else '(no content)'}")
+```
+
+## Update Subagent Settings
+
+```python
+# Change subagent's model
+client.agents.update(
+    'agent-a856f614-7654-44ba-a35f-c817d477dded',
+    model='claude-sonnet-4-20250514'
+)
+```