Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 32 additions & 0 deletions akf-trust-metadata/.cursor-plugin/plugin.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
{
"name": "akf-trust-metadata",
"displayName": "AKF — The AI Native File Format",
"version": "1.0.0",
"description": "EXIF for AI. Stamps every file agents generate with trust scores, source provenance, and compliance metadata. ~15 tokens of JSON that embed into 20+ formats.",
"author": {
"name": "HMAKT99",
"email": "akf@akf.dev"
},
"homepage": "https://akf.dev",
"repository": "https://github.com/HMAKT99/AKF",
"license": "MIT",
"keywords": [
"trust",
"metadata",
"provenance",
"compliance",
"ai-native",
"file-format",
"eu-ai-act",
"exif-for-ai"
],
"category": "developer-tools",
"tags": [
"trust",
"provenance",
"compliance",
"ai-safety",
"developer-tools"
],
"skills": "./skills/"
}
21 changes: 21 additions & 0 deletions akf-trust-metadata/LICENSE
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
MIT License

Copyright (c) 2026 HMAKT99

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
32 changes: 32 additions & 0 deletions akf-trust-metadata/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
# AKF — The AI Native File Format

EXIF for AI. Stamps every file your agents generate with trust scores, source provenance, and compliance metadata.

## What it does

When Cursor agents create or modify files, AKF stamps them with ~15 tokens of JSON carrying:
- **Trust scores** (0–1) based on source tier
- **Source provenance** — where the information came from
- **Compliance metadata** — EU AI Act, SOX, HIPAA, NIST readiness

The metadata embeds natively into 20+ formats — DOCX, PDF, images, Markdown, YAML, JSON, Python, TypeScript, Go, Rust, HTML — no sidecars.

## Quick start

```bash
pip install akf
akf stamp report.md --agent cursor --evidence "generated from user prompt"
akf inspect report.md
```

## Why

- **47% of developers** distrust AI-generated code (Cognition/Devin, 2026)
- **EU AI Act Article 50** takes effect August 2, 2026 — AI content must carry transparency metadata
- Every file agents generate should carry provenance

## Links

- [akf.dev](https://akf.dev)
- [GitHub](https://github.com/HMAKT99/AKF)
- [Demo](https://huggingface.co/spaces/HANAKT19/akf)
85 changes: 85 additions & 0 deletions akf-trust-metadata/skills/akf-stamp/SKILL.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,85 @@
---
name: akf-stamp
description: >-
Stamps files with AKF trust metadata — the AI native file format.
Use after creating or modifying any file to add trust scores, source
provenance, and compliance metadata. Use before modifying existing files
to check current trust context. Use when the user mentions trust,
provenance, compliance, EU AI Act, or metadata.
---

# AKF — The AI Native File Format

Every photo has EXIF. Every song has ID3. AKF is the native metadata format for AI-generated content — ~15 tokens of JSON that embed trust scores, source provenance, and compliance data into 20+ file formats.

## After creating or modifying files

Stamp every file you create or modify with trust metadata:

```bash
akf stamp <file> --agent cursor --evidence "<what you did>"
```

Evidence should describe what you did:
- `"generated from user prompt"` — new file from scratch
- `"refactored existing code"` — modification
- `"tests pass"` — after running tests
- `"docs reviewed"` — documentation
- `"bug fix, verified manually"` — fixes

## Before modifying existing files

Always check existing trust metadata first:

```bash
akf read <file>
akf inspect <file>
```

This shows who last modified the file, their evidence, and the trust score. Preserve or update this context.

## Classification

Apply security labels based on file location:

- `--label confidential` for files in `*/finance/*`, `*/secret/*`, `*/internal/*`
- `--label public` for README, docs, examples, open-source files
- Default is `internal`

## Compliance auditing

Check if files meet regulatory requirements:

```bash
akf audit <file> --regulation eu_ai_act # EU AI Act Article 50
akf audit <file> --regulation hipaa # HIPAA
akf audit <file> --regulation sox # SOX
akf audit <file> --regulation nist_ai # NIST AI RMF
```

## Trust scoring

Trust scores range from 0 to 1, based on source tier:

| Tier | Score | Examples |
|------|-------|----------|
| T1 | 0.95 | SEC filings, court records, official APIs |
| T2 | 0.85 | Reuters, peer-reviewed journals |
| T3 | 0.70 | News articles, Wikipedia |
| T4 | 0.50 | Blog posts, forums |
| T5 | 0.30 | Unverified AI output |

Human review promotes AI content from T5 (0.30) to T3+ (0.70+).

## Install

```bash
pip install akf
```

## Links

- Website: https://akf.dev
- GitHub: https://github.com/HMAKT99/AKF
- npm: `npm install akf-format`
- Demo: https://huggingface.co/spaces/HANAKT19/akf