- Upstream:
governex/agent-action-receipts-vectors - Pinned commit:
6e31f1fabe0f5f6de511c5821bdf8b924d8aaa2a - Draft profile:
draft-sahu-agent-action-receipts-01 (vectors 01-18 are unchanged from the -00 suite; 19-21 and the head assertion exercise -01 additions) - Verifier:
openpoc/action_receipt_compat_v01.py - Upstream verifier execution: disabled; CI reads only the pinned manifest and vector data.
- Prior
01–18/-00report: remains pinned and unchanged.
Checked 7 October 2026: the IETF Datatracker lists only draft-sahu-agent-action-receipts-00, dated 16 August 2026. Its published Implementation Status does not name Aleksei Safonov or T-Trace. This report's 18/18 result is pinned repository-level interoperability evidence for the upstream -01 vector/profile state; it is not evidence that an IETF -01 revision has been published, adopted, or endorsed.
18/18 checks agree; 0 disagree; 0 unsupported.
| Vector | Exp. sig fail | Obs. sig fail | Exp. chain | Obs. chain | Exp. step repeat | Obs. step repeat | Exp. seq gap | Obs. seq gap | Exp. seq repeat | Obs. seq repeat | Result |
|---|---|---|---|---|---|---|---|---|---|---|---|
01-basic-chain.jsonl |
— | — | — | — | — | — | — | — | — | — | AGREE |
02-optional-actor.jsonl |
— | — | — | — | — | — | — | — | — | — | AGREE |
03-extension-members.jsonl |
— | — | — | — | — | — | — | — | — | — | AGREE |
04-params-shapes.jsonl |
— | — | — | — | — | — | — | — | — | — | AGREE |
10-tampered-value.jsonl |
1 | 1 | 2 | 2 | — | — | — | — | — | — | AGREE |
11-reordered.jsonl |
— | — | 2 | 2 | — | — | — | — | — | — | AGREE |
12-interior-deleted.jsonl |
— | — | 2 | 2 | — | — | — | — | — | — | AGREE |
13-duplicated-line.jsonl |
— | — | 3 | 3 | 3 | 3 | — | — | — | — | AGREE |
14-head-truncated.jsonl |
— | — | — | — | — | — | — | — | — | — | AGREE |
15-genesis-with-prevhash.jsonl |
— | — | 1 | 1 | — | — | — | — | — | — | AGREE |
16-missing-prevhash.jsonl |
— | — | 2 | 2 | — | — | — | — | — | — | AGREE |
17-bad-key-encoding.jsonl |
1 | 1 | — | — | — | — | — | — | — | — | AGREE |
18-tampered-extension.jsonl |
— | — | 2 | 2 | — | — | — | — | — | — | AGREE |
19-replayed-step-id.jsonl |
— | — | — | — | 4 | 4 | — | — | — | — | AGREE |
20-seq-gap.jsonl |
— | — | — | — | — | — | 3 | 3 | — | — | AGREE |
21-seq-repeat.jsonl |
— | — | — | — | — | — | — | — | 3 | 3 | AGREE |
| Assertion | Log | Expected | Observed | Result |
|---|---|---|---|---|
30-head-assertion.json |
01-basic-chain.jsonl |
match |
match |
AGREE |
30-head-assertion.json |
14-head-truncated.jsonl |
mismatch |
mismatch |
AGREE |
19-replayed-step-idisolates semantic record repetition: signatures and raw-octet linkage remain valid, while repeatedstep_idis rejected.20-seq-gapand21-seq-repeatdistinguish a missing signed position from a reused/regressed signed position inside the optional per-chain sequence profile.30-head-assertion.jsondetects presentation of the truncated vector relative to a particular signed external head state.
These mechanisms separate record repetition, recorder-issued positional anomalies, and head truncation. They do not by themselves prove that every real effect was captured. In particular:
- a repeated real-world effect with a fresh
step_idand the next validseqremains record-valid unless a stable effect identity, authorization nonce, request digest, or idempotency binding is also signed; - a
seqgap proves a gap in the recorder's signed numbering, not an unrecorded effect, unless sequence allocation is itself non-bypassable and occurs before the effect; - a head assertion proves consistency against that assertion; preventing equivocation between multiple valid assertions still requires witness, gossip, transparency, or another monotonic external state.
Agreement is interoperability evidence only. It does not:
- prove the draft correct or complete;
- prove real-world capture completeness;
- prove that the head signer cannot equivocate;
- turn
step_idorseqinto effect-level anti-replay binding; - constitute endorsement in either direction.