What is in llamafarm's system prompt?
llamafarm's full system prompt: 1 version, 4,600 characters. Audited against AISPA.
The full text of 1
prompt is reproduced below,
4,600 characters in all, each read
instruction by instruction against the eight
AISPA dimensions.
Nothing was flagged as working against the person on
the other end.
4600 characters
---
name: generate-subsystem-skills
description: Generate specialized skills for each subsystem in the monorepo. Creates shared language skills and subsystem-specific checklists for high-quality AI code generation.
allowed-tools: Read, Grep, Glob, Write, Edit, Task, Bash
---
# Generate Subsystem Skills
This skill analyzes each subsystem in the LlamaFarm monorepo and generates specialized Claude Code skills for security, performance, and language-specific best practices.
## Usage
```
/generate-subsystem-skills
```
---
## What Gets Generated
### Shared Language Skills (4)
- `python-skills/` - Used by: server, rag, runtime, config, common
- `go-skills/` - Used by: cli
- `typescript-skills/` - Used by: designer, electron
- `react-skills/` - Used by: designer
### Subsystem-Specific Skills (8)
- `cli-skills/` - Cobra, Bubbletea patterns
- `server-skills/` - FastAPI, Celery, Pydantic patterns
- `rag-skills/` - LlamaIndex, ChromaDB patterns
- `runtime-skills/` - PyTorch, Transformers patterns
- `designer-skills/` - TanStack Query, Tailwind, Radix patterns
- `electron-skills/` - Electron IPC, security patterns
- `config-skills/` - Pydantic, JSONSchema patterns
- `common-skills/` - HuggingFace Hub patterns
---
## Generation Process
### Step 1: Read Registry
Load subsystem definitions from [subsystem-registry.md](subsystem-registry.md).
### Step 2: Generate Shared Language Skills
Launch sub-agents IN PARALLEL to generate:
1. **Python Skills Agent** - Analyze Python subsystems (server, rag, runtime, config, common), identify ideal patterns, generate `python-skills/`
2. **Go Skills Agent** - Analyze CLI subsystem, identify ideal Go patterns, generate `go-skills/`
3. **TypeScript Skills Agent** - Analyze designer and electron, identify ideal TS patterns, generate `typescript-skills/`
4. **React Skills Agent** - Analyze designer, identify ideal React 18 patterns, generate `react-skills/`
### Step 3: Generate Subsystem Skills
Launch sub-agents IN PARALLEL for each subsystem:
For each subsystem, the agent should:
1. Read the subsystem's dependency files (package.json, pyproject.toml, go.mod)
2. Analyze code patterns using Grep and Read
3. Generate SKILL.md that links to shared language skills
4. Generate framework-specific checklist files
5. Write all files to `.claude/skills/{subsystem}-skills/`
### Step 4: Report Summary
After all agents complete, report:
- Number of skills generated
- Total files created
- Any errors encountered
---
## Sub-Agent Prompt Templates
### For Shared Language Skills
```
You are generating a shared {LANGUAGE} skills directory for Claude Code.
Analyze these subsystems that use {LANGUAGE}:
{SUBSYSTEM_PATHS}
Your task:
1. Read key files to understand patterns used
2. When patterns vary, document the IDEAL approach (not inconsistencies)
3. Reference industry best practices
4. Generate files in .claude/skills/{LANGUAGE}-skills/
Files to generate:
- SKILL.md (overview, ~100 lines)
- patterns.md (idiomatic patterns)
- error-handling.md
- testing.md
- security.md
- {additional language-specific files}
Each checklist item should have:
- Description of what to check
- Search pattern (grep command)
- Pass/fail criteria
- Severity level
```
### For Subsystem Skills
```
You are generating subsystem-specific skills for {SUBSYSTEM} in Claude Code.
Directory: {PATH}
Tech Stack: {TECH_STACK}
Links to: {SHARED_SKILLS}
Your task:
1. Read dependency files and key source files
2. Identify framework-specific patterns
3. Generate SKILL.md that links to shared language skills
4. Generate framework-specific checklists
Files to generate:
- SKILL.md (overview with links to shared skills)
- {framework}.md for each framework used
- performance.md (subsystem-specific optimizations)
Remember: Document IDEAL patterns, not existing inconsistencies.
```
---
## Key Principle
**Prescribe ideal patterns** - When the codebase has inconsistent patterns, the generated skills should document the BEST practice according to industry standards, not codify existing inconsistencies.
---
## Output Location
All skills are written to `.claude/skills/` with this structure:
```
.claude/skills/
├── python-skills/ # Shared
├── go-skills/ # Shared
├── typescript-skills/ # Shared
├── react-skills/ # Shared
├── cli-skills/ # Subsystem
├── server-skills/ # Subsystem
├── rag-skills/ # Subsystem
├── runtime-skills/ # Subsystem
├── designer-skills/ # Subsystem
├── electron-skills/ # Subsystem
├── config-skills/ # Subsystem
└── common-skills/ # Subsystem
```
Questions about llamafarm's system prompt
Does llamafarm's system prompt contain instructions that work against the user?
No. Nothing in llamafarm's system prompt was flagged as working against the person the product is talking to. That is a clean result across all eight AISPA dimensions, not an absence of checking — the full text was read instruction by instruction.
How long is llamafarm's system prompt?
4,600 characters across 1 prompt on this page. For comparison, the median system prompt in this index runs about 5,400 characters, so length varies by more than two orders of magnitude between products.
How many versions of llamafarm's system prompt are on record?
1. Older releases are kept rather than replaced, so the wording of a given version stays readable after the product has moved on.
Where did this llamafarm system prompt come from?
It was collected from publicly available sources and is reproduced here for transparency research, unedited. This site does not extract prompts from products itself.
How was llamafarm's system prompt audited?
Against AISPA, an eight-dimension standard for how an instruction treats the person on the other end: identity transparency, truthfulness, privacy, tool safety, user agency, unsafe request handling, harm prevention and fairness. This audit was ai audit. The method is described in the paper behind the standard.