Claude-skill-registry fleet-agent

Context-aware development assistant for AgenticFleet with auto-learning and dual memory (NeonDB + ChromaDB). Handles development workflows with intelligent context management.

install

source · Clone the upstream repo

git clone https://github.com/majiayu000/claude-skill-registry

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/fleet-agent" ~/.claude/skills/majiayu000-claude-skill-registry-fleet-agent && rm -rf "$T"

manifest: skills/data/fleet-agent/SKILL.md

Fleet Agent

A context-aware development assistant for AgenticFleet that maintains persistent memory across sessions using a hybrid NeonDB + ChromaDB architecture.

Memory Architecture

Dual Storage

ChromaDB (Semantic): Skills, patterns, code snippets with embedding-based search
NeonDB (Structured): Sessions, users, analytics, skill metadata with SQL queries

Context Layers

Core Memory (
```
.fleet/context/core/
```
): Always loaded
- ```
project.md
```
  : Architecture, conventions, tech stack
- ```
human.md
```
  : User preferences, communication style
- ```
persona.md
```
  : Agent guidelines, tone
Topic Blocks (
```
.fleet/context/blocks/
```
): Loaded on demand
- ```
project/
```
  : commands, conventions, gotchas, architecture
- ```
workflows/
```
  : git, review
- ```
decisions/
```
  : ADRs
Skills (ChromaDB + NeonDB): Semantic + structured patterns

Usage Examples

Learn a Pattern

/fleet-agent learn --name "add_dspy_agent" --category "agent" --content "Create agent via AgentFactory with DSPyEnhancedAgent wrapper..."

Recall Information

/fleet-agent recall "DSPy typed signatures"
/fleet-agent context "add a new agent for web search"

Analyze Code

/fleet-agent analyze src/agents/coordinator.py

Session Management

/fleet-agent session start
/fleet-agent session status
/fleet-agent session summary "Completed agent creation workflow"

Commands

Command	Description
`learn --name <name> --category <cat> --content <code>`	Save pattern to both databases
`recall <query>`	Search NeonDB + ChromaDB
`context <task>`	Load relevant context blocks
`analyze <file>`	Analyze code structure
`session start`	Start new session
`session status`	Show current session
`session summary <text>`	Save session summary
`stats`	Show development metrics

Auto-Learning

Automatically extracts and saves patterns after successful task completion with detailed code examples:

name: pattern_add_dspy_signature
category: dspy
description: How to create a DSPy signature with TypedPredictor
implementation: |
  class TaskAnalysisOutput(BaseModel):
      complexity: Literal["low", "medium", "high"]

  class TaskAnalysis(dspy.Signature):
      task: str = dspy.InputField(desc="Task to analyze")
      analysis: TaskAnalysisOutput = dspy.OutputField()

Implementation

Main script:

.fleet/context/scripts/fleet_agent.py

Invocation:

uv run python .fleet/context/scripts/fleet_agent.py <command>

Dependencies:

neon_memory.py

chroma_driver.py

memory_loader.py