Model Bundle Integrity Manifest
Refuses to write if the directory holds a file you did not declare, so the manifest describes the exact topology.
$29
Works with the AI Agents you already use
Model Bundle Integrity Manifest
Example session with this skill installed
Generate a manifest for my model directory at ./dist/llama-v3 with roles for the transformer and tokenizer, pinned to onnxruntime on cpu.
- Read your context and instructions
- Compiled the model bundle integrity
Generated manifest for ./dist/llama-v3. Every file has been hashed and pinned to the specified runtime.
Manifest Summary
{
"schemaVersion": "1.0.0",
"engineVersion": "1.4.0",
"manifestSha256": "8f3e27a1c92b45d6e81f72c9a871d3e192837465afbc9021034857261a938c42",
"runtime": {
"backend": "onnxruntime",
"device": "cpu",
"dtype": "float32"
},
"models": [
{
"role": "transformer",
"file": "model.onnx",
"bytes": 452083120,
"sha256": "a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2"
},
{
"role": "tokenizer",
"file": "tokenizer.json",
"bytes": 210432,
"sha256": "f0e9d8c7b6a5f4e3d2c1b0a9f8e7d6c5b4a3f2e1d0c9b8a7f6e5d4c3b2a1f0e9"
}
]
}
Next steps
- Copy the
manifest.jsoninto your./dist/llama-v3directory. - Add
const { verifyBundleDirectory } = require('./scripts/verify-bundle.mjs')to your service entry point. - Record the
manifestSha256value in your deployment logs to verify consistency across your cluster.
Connects securely to your tools. The creator never sees your data.
About this skill
Model weights are the one artifact most deployments never verify after copying. A swapped file, a download truncated by a dropped connection, a weight replaced through a symbolic link or quietly updated to another quantization loads without complaint and changes every answer the service gives.
Manifest (make-manifest.mjs): writes manifest.json from the directory you intend to ship: engine version, a runtime map (backend, device, dtype, versions), and for every model its role, plain file name, exact byte count, SHA-256, source and license. Refuses to write if the directory holds a file you did not declare, so the manifest describes the exact topology.
Verifier (verify-bundle.mjs, library and CLI): the directory and every file must be real (lstat, realpath equal, no symbolic links or junctions); every file is stat'ed before and after it is streamed through SHA-256 and must be byte-identical and unchanged in inode, size, mode and timestamps; the directory listing must be exactly the manifest plus the listed files; roles must appear in the order your build expects; runtime fields you pin must match; unknown manifest keys fail. Returns the manifest digest so your health endpoint can prove which weights are loaded; exits 1 with MODEL_INTEGRITY: <reason> otherwise.
6 tests: intact bundle, generator provenance fields and undeclared-file refusal, byte-count and byte-flip and extra-file and missing-file failures, role order and duplicate roles and path escapes and extension limits and runtime pins, symbolic-link redirection, and the two CLIs end to end.
reference/manifest-format.md documents the format and each check. Node 20+, no packages, no network.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Requires Node 20+. Cannot protect against attackers with write access to both manifest and model files. Symbolic link detection tests require specific OS privileges on Windows.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
- Software
Node.js 20+