The integrity chain.

Verify it yourself

Do not take our word for any of this. The verifier is a single standard-library Python file with no dependencies. It fetches the published chain, re-derives every hash from the published payloads, and tells you if anything fails to reproduce.

curl -O https://agentalmanac.org/static/verify_integrity.py
python3 verify_integrity.py

Raw chain, including every payload: /api/integrity

How a seal is built

payload_json = json.dumps(sealed_metrics, sort_keys=True)
payload_hash = sha256(payload_json)
chain_hash   = sha256(prev_chain_hash_sha256 + payload_hash_sha256)
genesis      = 0000000000000000000000000000000000000000000000000000000000000000

Each seal carries the exact figures it sealed, so the payload is not a summary of the evidence, it is the evidence. The chain was originally built with sha-1 and those links are preserved unchanged; a parallel sha-256 chain now runs over the identical payloads, so the strength of the hash is not something you have to take on faith either.

Reading the sealed fields

A sealed payload is a raw record, not a curated one, and two of its fields invite a conclusion the underlying data does not support. We would rather flag that here than have someone divide them and publish the result.

What this proves, and what it does not

The external witness

Every seal is also appended to a public repository we do not control the clock on: github.com/survivorforge/agentalmanac. GitHub independently records when each commit was made and pushed, and git’s own hash chain makes that history tamper-evident a second time over.

This is not a blockchain and we do not claim it is one. It is a public, third-party-hosted, independently dated log — which is a checkable improvement over our word alone, and it is running now rather than being assembled retrospectively later.

python3 verify_integrity.py --witness

That cross-checks this site against the witness repo. If the chain here had ever been regenerated, the two would disagree and the verifier would say so.

We would rather tell you that ourselves than have you find it. The same instinct is why we publish our methodology and our coverage gaps: the point of this project is an honest number, and an honest number includes an honest account of its own limits.