[Web-bot-auth] Re: Interest in the human-principal layer above bot authentication

[Songbo Bu <[e-mail]>]

Nenad, all,

This is a useful post-execution evidence layer, but I would keep it
separate from the identity and authority layers rather than letting an
audit-chain result absorb their claims.

In the principal-binding model, I would initially map the contract
across five rows:

- C-005: which protected fields bind the record to the same action or
governed event;
- C-007: which recorder, witness, key, retained prefix, and evolution
boundary authenticate the evidence;
- C-010: how mission, principal, authorization, runtime, and admission
records compose without changing their meanings;
- C-011: the constrained result a consumer may accept after verification; and
- C-012: whether the record is pre-execution authority, post-execution
attribution, execution evidence, or audit enforcement.

The key boundary is that tamper-evident does not mean true, complete,
authorized, independently witnessed, or successfully executed. A
signing key can anchor attribution under an explicit enrollment and
key-control rule, but it does not establish mission authority or human
consent by itself.

I checked the public coordinates you supplied. PR 3004 was open and
unmerged at head:

377f8d260ded5b6854871b2ce3c73621ffcaef1d

The PR names the vector implementation pin:

notboatanchor/gif@e1f02a95506e81e7766c3ba3a684ecad7cfff12f

I have not converted the PR's reported 23/23 result into independent
evidence. The attached note records a bounded reproduction and mapping
plan, including retained-prefix verification, truncation, fabricated
evolution boundaries, witness disagreement, downgrade to unattested
records, canonicalization ambiguity, algorithm transition, and
conflicting continuations.

Before adding a protocol mapping to the draft, the most useful
evidence would be:

1. a manifest for the exact vector and verifier files;
2. the expected constrained result for each negative case, not only pass/fail;
3. the witness independence and disagreement rule;
4. an algorithm-transition vector that preserves the prior prefix
without rewriting it; and
5. the exact command and environment needed to reproduce the corpus.

If you publish those pins and artifacts in this thread, I can run the
corpus independently and return a C-005/C-007/C-010/C-011/C-012
mapping with every scenario labeled measured, not-run, or
not-applicable.

Best,
Songbo