Waiver — PTV-SCF-0001 P2 · G-P2-2 and G-P2-3

rank: waiver waives: [G-P2-2, G-P2-3] verdict_source: docs/decisions/2026-08-13-ptv-scf-0001-p2-verification-round.md discharge: P3 · Verification harness — F-22, F-23, F-24, F-26, F-27, F-28, F-29, F-30, F-31 expires: void if P3 closes without discharging F-23, F-24 and F-26

Waiver — PTV-SCF-0001 P2 · G-P2-2 and G-P2-3

Date: 2026-08-13 Status: open Supersedes: none Superseded-by: none — current Shape: docs/decisions/2026-08-13-acceptance-gate-waiver-shape.md

Authored on explicit operator instruction (“define the waiver shape first, then waive both gates”), which is the R-W4 condition. No agent inferred this waiver from an approval of adjacent work.

1 · What failed, verbatim

G-P2-2, as declared at docs/decisions/2026-08-13-ptv-scf-0001-p2-open.md:164:

A session using only the fetched preamble performs equivalently to one given the pasted §1 block. Equivalent means: same law cited, same refusals predicted, no governance claim present in one and absent in the other.

Verdict, from the merged round, Part A: FAIL.

Both cited L1–L7 and W1–W4 and both predicted the .github/workflows/ refusal. They diverge on the gate’s third clause — no governance claim present in one and absent in the other. The pasted arm found that console is a git submodule, concluded that a rocky-hq-scoped verb can only move its pointer and cannot write console source at all, and refused to name a verb because no catalogue is on disk. The fetched arm made none of those claims.

G-P2-3, as declared at …p2-open.md:165:

No skill requires out-of-band context to work. FAIL if any skill’s correct operation depends on a fact not reachable from the skill text plus the fetched artefacts. Asserted per skill, not in aggregate.

Verdict: FAIL, 8 of 8.

The recurring one is fatal to the phase objective: no skill says where a repo’s phase state lives. Four runs independently guessed MILESTONES.md, and one noted it also had to decide unaided whether that file counts as a source or as a projection under L5.

Neither verdict is restated more kindly here than the round recorded it. Both gates read FAIL and continue to read FAIL after this waiver.

2 · Why closing anyway is legitimate

Not because the gates were too strict. They were written correctly and they caught real defects — G-P2-3 in particular found a gap that a friendlier gate would have missed entirely.

P2’s deliverables stand on their own evidence:

The eight skills exist, load cold, and reach their declared terminal states. G-P2-1 passed on 8/8 runs — four artefacts, four correct refusals, including one that declined to route around a privileged-path block when routing around it would have been easy.
The three machine artefacts carry every field the phase specified. G-P2-4 passed, re-run at round time.
Gate G4, asserted by CI since the SDD was written and implemented by nobody, now runs and fails correctly on a seeded orphan. G-P2-5 passed.
The 35-slug allocation ledger is complete and matches the estate’s own computed population. G-P2-6 passed as delivered.
The merged phase-cycle skill — the specific anti-shape task 5 forbids — is retired.

What the two failed gates establish is that the fetch path is not yet trustworthy, and the skills under-specify where repo state lives. Both are real, both are enumerated with targets, and neither undoes the four gates that passed. The phase produced what it set out to produce; it also discovered that its objective is one layer deeper than it assumed. Closing on that finding, with the debt named, is more honest than holding a phase open indefinitely while the next phase — whose literal name is Verification harness — is the phase equipped to fix it.

3 · Scope test (R-W2)

Answered explicitly, and the honest answer is not uniform across the two gates.

G-P2-3 · passes the scope test cleanly. The failure is in eight specific skill files authored in P2. Each gap has a named fix — say where phase state lives, promote verb-schema fetches into ## Load first, add the already-admitted exit to petrova-onboard, enumerate the vocabularies the skills mandate. Fixing them fixes the gate. Nothing about it recurs structurally.

G-P2-2 · passes only conditionally, and this is the weak point of this waiver. The cause is the fetch transport returning a summary rather than the artefact, and possibly refusing the artefact outright as a suspected injection. That condition is not P2-specific — it degrades every phase that relies on fetched law, which is every phase from here. Under a strict reading of R-W2 that is the propagating condition the rule says must be fixed rather than waived.

It is waived anyway, on one basis: the expiry in §5 binds it to P3. A propagating condition with a hard discharge deadline is a different object from a propagating condition waived open-endedly, which is what R-W2 exists to prevent. If P3 closes without discharging F-23 and F-24, this waiver voids and P2 is retrospectively closed on a lapsed instrument — a consequence stated here so nobody has to discover it later.

Recording the tension rather than resolving it in the waiver’s favour is the point of the section. A reader who thinks R-W2 should have blocked G-P2-2 has been given everything needed to make that argument.

4 · What carries

Waived gate	Underlying work	Friction IDs	Target
G-P2-2	The fetch path must deliver the artefact, not a paraphrase of it, and must not be intermittently refused as an injection attempt. Serve it in a form a fetcher returns literally, or publish a fetch-and-verify recipe using the `Law fingerprint` as the check.	F-23, F-24	P3
G-P2-2	The preamble contradicts itself on W1 — line 35 forbids citing meta-rules by number, line 60 instructs it. One is wrong.	F-22	P3
G-P2-3	Say where phase state lives; the mandatory load (`contract.yaml`) carries none.	F-26	P3
G-P2-3	`petrova-onboard`: add the already-admitted exit, and name where the control-plane registry lives — a consumer’s own `registry.yaml` is a wrong-answer trap.	F-27	P3
G-P2-3	Promote per-verb schema fetches from `## Reference` to `## Load first`.	F-28	P3
G-P2-3	`.petrova/brand.yaml` is absent in consumers and in the control plane itself; ship it or state the fallback.	F-29	P3
G-P2-3	Enumerate the vocabularies the skills mandate — severity scales, legal milestone states.	F-30	P3
G-P2-3	Exercise the `commands/` slash-command invokers, still unrun.	F-31	P3
both	`close_phase` cannot express a waiver; “merged” is undefined as a testable property.	F-37, F-38	P3

Every ID above already exists in a merged record — the round for F-22…F-35, the close-refusal finding for F-36/F-37, the shape record for F-38. No row invents work to look thorough.

5 · Discharge and expiry (R-W3)

Discharged when F-22, F-23, F-24, F-26, F-27, F-28, F-29, F-30 and F-31 are done and evidenced in P3, and a P3 verification round records G-P2-2 and G-P2-3 as re-evaluated PASS against fresh runs. Not by assertion, and not by the agent that does the work.

Void if P3 closes without discharging F-23, F-24 and F-26 — the three that are load-bearing rather than incidental. On voiding: G-P2-2 and G-P2-3 revert to FAIL, and P2 stands closed on a lapsed waiver, which is itself a finding to be raised at P3’s close rather than a matter of interpretation.

The other six items may slip to P4 without voiding this waiver. That asymmetry is deliberate: the fetch path and the missing phase-state pointer are what make the law unusable in practice; a missing severity vocabulary is an irritation.

6 · What this does not license

It does not license reading either gate as passed. Any close record, any projection, any summary citing this waiver says “closed with G-P2-2 and G-P2-3 waived”. It never says the gates are green, and it never omits them.
It does not license waiving these gates again in P3. P3 inherits the work, not the exemption. A second waiver of the same gates would be the permanent excuse R-W3 exists to prevent, and R-W2 would refuse it outright.
It does not license pasting the preamble as a workaround. The pasted arm outperformed the fetched arm in one comparison; that is a finding about the fetch path, not a reversal of the phase’s objective. Skills continue to fetch.
It does not license closing P2 on the remaining preconditions. This waiver satisfies precondition 3 alone. Preconditions 1, 2 and 4 stand on their own evidence and the close must still evaluate them.
It does not license treating the rocky-hq findings (F-32) as this repo’s to absorb. They belong in that repo’s ledger under L4, waiver or no waiver.

References

docs/decisions/2026-08-13-ptv-scf-0001-p2-verification-round.md — the merged round carrying both FAIL verdicts and the friction IDs.
docs/decisions/2026-08-13-ptv-scf-0001-p2-open.md:164-165 — the gates as declared.
docs/findings/20260813-2047-p2-close-refused-at-precondition-3.md — the refusal this answers, and F-36/F-37.
docs/decisions/2026-08-13-acceptance-gate-waiver-shape.md — R-W1…R-W4.

Sign-off

Subagent: PTV-SCF-0001 P2 C5 (session 2026-08-13)
Human: ☑ (proxy) countersign — waives G-P2-2 and G-P2-3 for PTV-SCF-0001 P2 only, on the carry in §4 and the discharge condition in §5. Does not assert that either gate passed.
- Countersigned by human:devarno on 2026-08-13, by explicit directive in session (“define the waiver shape first, then waive both gates”). Ticked by the agent as scribe, not as signatory: the human act is the instruction to waive, naming what is waived, which is what R-W4 requires.