Hermes Figma Prompt Hub Server

CI License: MIT Sponsor OpenCollective

Figma-to-MCP prompt management for Hermes workflows.

Hermes Figma Prompt Hub is an open-source scaffold for designing, versioning, validating, and eventually serving structured prompt assets from Figma into Hermes-compatible MCP workflows.

The project keeps Figma-specific integration work outside the Hermes core repository while giving teams a clear place to build prompt tooling, validation, importer logic, examples, and automation.

Upstream

Hermes core repository:

https://github.com/NousResearch/hermes-agent

Local upstream remote:

git remote add upstream [email protected]:NousResearch/hermes-agent.git

This project remains an external integration repository so Figma-specific code can evolve outside the Hermes core tree.

Status

This repository is currently an integration scaffold.

• Ready: prompt schema, example prompt, validation scripts, Figma layer contract, CI, contributor automation, sponsorship docs.
• In progress: Figma frame importer and MCP server implementation.
• Not ready yet: production Figma sync, full MCP tool server, package release.

What It Provides

• A structured JSON prompt format for system, developer, user template, variables, guardrails, tools, output format, examples, changelog, and eval checks.
• A Figma layer naming contract for frames named PROMPT/<id>.
• Validation scripts for prompts, schema consistency, prompt catalog generation, and Figma contract alignment.
• Prompt pattern analysis and prompt security audit reports.
• GitHub automation for CI, code review, contributor onboarding, issue health, prompt catalog updates, and collaboration scouting.
• Documentation and issue templates for contributors, sponsors, and maintainers.

Use Cases

• Design prompts visually in Figma and map layers into structured prompt JSON.
• Maintain prompt versions with guardrails, examples, output contracts, and eval checks.
• Prepare prompt assets for Hermes or MCP-compatible tools.
• Build a shared prompt hub for design, product, and engineering workflows.
• Invite collaborators around Figma API, MCP, prompt evaluation, docs, and automation.

Quickstart

git clone https://github.com/jozrftamson/hermes-figma-prompt-hub.git
cd hermes-figma-prompt-hub
python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
python scripts/validate_repo.py

Validate one prompt:

python scripts/validate_prompt.py prompts/raw/nous-central-v1.json

Regenerate the prompt catalog:

python scripts/generate_prompt_catalog.py

Create a scaffold in another directory:

python install/scaffold.py /path/to/figma-hermes-prompt-hub

Prompt Format

Prompt files live under prompts/raw/*.json and must match prompts/schema/prompt.schema.json.

Core fields:

• id, version, status, category, tags
• system, developer, user_template
• variables, context_sources, tools
• guardrails, constraints, style_rules
• output_format
• few_shots
• changelog
• eval.must_include, eval.must_not_include, eval.checks

Example:

{
  "id": "nous-central-v1",
  "version": "0.2.0",
  "status": "active",
  "category": "general-assistant",
  "variables": ["context", "task", "format"],
  "output_format": {
    "type": "json",
    "instructions": "Return valid JSON with summary, reasoning_notes and next_actions."
  },
  "guardrails": ["Keine erfundenen Fakten", "Bei Unsicherheit klar markieren"],
  "eval": {
    "must_include": ["summary", "next_actions"],
    "must_not_include": ["Great question", "As an AI"],
    "checks": [
      {
        "name": "valid_json_output",
        "type": "json_schema"
      }
    ]
  }
}

See the full example at prompts/raw/nous-central-v1.json.

Figma Contract

Create Figma frames named PROMPT/<id>. Text layers should use the naming contract in prompts/figma-layer-contract.txt.

Current layer names:

00_system
01_developer
02_user_template
03_output_format
04_tool_policy
05_context_source_<name>
10_guardrail_<n>
11_constraint_<n>
12_style_rule_<n>
20_example_in_<n>
21_example_out_<n>
22_example_note_<n>
30_variable_<name>
40_test_case_<n>
50_expected_output_<n>
80_changelog_<version>
90_eval_must_include_<n>
91_eval_must_not_include_<n>
92_eval_check_<n>

Fixture data for importer development lives under evals/cases.

MCP Direction

The planned MCP server should expose tools for:

• listing available prompt IDs
• validating prompt files
• exporting prompt content as structured JSON
• later syncing or importing Figma frames

Hermes should consume this as an external MCP integration instead of adding Figma-specific code to the Hermes core tree.

Example configuration shape:

{
  "mcp_servers": {
    "figma-prompt-hub": {
      "command": "python",
      "args": ["/absolute/path/to/hermes-figma-prompt-hub/mcp/server.py"],
      "env": {
        "FIGMA_ACCESS_TOKEN": "${FIGMA_ACCESS_TOKEN}",
        "FIGMA_FILE_KEY": "${FIGMA_FILE_KEY}"
      }
    }
  }
}

Automation

The repository includes automation for development quality and community growth:

• CI validation
• automated code review with severity policy, labels, inline comments, and PR score
• Figma contract check
• prompt catalog generation
• automation failure diagnostics
• good-first-issue seeding
• stale issue pings without automatic closing
• contributor recognition
• contributor scout reports
• monthly sponsor and contributor updates
• CODEOWNERS routing

Useful local commands:

python scripts/validate_repo.py
python scripts/check_figma_contract.py
python scripts/analyze_prompt_patterns.py
python scripts/security_prompt_audit.py
python scripts/import_prompt_text.py local-prompt.txt --id my-prompt
python scripts/generate_prompt_catalog.py
python scripts/code_review.py --base origin/main
python scripts/diagnose_automation_failures.py --limit 30
python scripts/scout_collaborators.py

Roadmap

Near-term work:

1. Implement the Figma importer for PROMPT/<id> frames.
2. Implement the first MCP server tools.
3. Add more prompt examples and eval fixtures.
4. Improve prompt semantic review.
5. Prepare a stable schema release.

See ROADMAP.md.

For a deeper technical overview, see docs/architecture.md. For first-time setup, see docs/getting-started.md. For project-board setup, see docs/project-board.md.

Contributing

Good first areas:

• Figma API importer
• MCP server tools
• prompt examples
• eval fixtures
• validation and review automation
• documentation

Start with CONTRIBUTING.md, COLLABORATION.md, and open issues labeled good first issue or help wanted.

Sponsoring

OpenCollective is the recommended funding path for transparent project support:

https://opencollective.com/hermes-figma-prompt-hub

GitHub Sponsors:

https://github.com/sponsors/jozrftamson

See SPONSORING.md.

License

MIT. See LICENSE.