TASK-120 — a KR's current is an assertion, and the payload says so

[ranjiao]

What was wrong

phase/<NNN>-linkage.md carries target and current per KR, both hand-written, and nothing derived, checked or aged them. The row's title says the edges are never folded into progress; the measurement says worse — the numbers were wrong in both directions on the same day:

KR	target	current	read as	actually
P-O1.1	1.0	0.0	0%	all four linked tasks closed
P-O2.2	0.0	0.0	met	TASK-094 measured 13 row splits and 87 header resolutions still live

Six of the eight phase KRs drive a count to zero, and the register template writes current: 0, so a drive-to-zero KR reads as achieved on the day it is written.

What this adds

Three additive blocks per KR, computed once in bin/lib § kr_progress_provenance and emitted by both perry-goals list --json and perry-state --json § linkage — one derivation, two payloads, so they cannot come to disagree.

• current_provenance — asserted / unasserted, when, at what granularity (asserted_scope: "register" — the date belongs to the register, not the KR), and measured: false. That last is true of every current Perry publishes, because no tool re-runs a KR's metric. It is emitted rather than implied so a consumer showing "measured" is contradicted by the payload rather than by a docstring.
• current_staleness — whether a linked task changed state after the number was asserted, naming which and how it moved. evaluated is carried apart from stale so nothing moved cannot be read as there was nothing to ask.
• linked_task_completion — {total, done, dropped, open, unknown}. A count of tasks, beside current and never inside it.
• conformance.krs_with_stale_current, and a perry-state warning naming the KRs.

The trap, and where the line is drawn

Nothing here computes current. "TASK-091 is closed" does not establish that the count its KR asks about is zero — only re-running the count does — and a ratio presented as current would be the fabricated measurement perry/OKR.md § Operating Principles forbids in its first line.

The linked-task tally is therefore emitted beside current, in a different unit, under its own name. On P-O1.1 the payload now carries current: 0.0, target: 1.0 and 4 of 4 tasks done simultaneously — it states the contradiction and resolves none of it. There is deliberately no conformance entry for "current disagrees with its tasks": that judgement needs the metric re-run.

Contract

perry-goals/list → 2.1. Four keys added, none removed or retyped, every one documented in schema/goals-list-contract.md in the same change.

• tests/test_contract_key_parity.py, perry-goals/list: before 0 documented-not-emitted / 5 emitted-not-documented; after 0 / 5 — the same five keys, which are TASK-131's. None of the new fields hides inside that number. Baseline re-recorded only because the contract-id key renamed.
• tests/test_contract_invariance.py green; its baseline is deliberately not re-recorded (see Notes).

Verification

• python3 tests/parallel -j 4: 66 modules · 1951 tests, 1 red module — test_diagnose, the same 2 failures as the baseline taken in this worktree before any edit (65 · 1930, same module, same 2). That is TASK-126's, not this row's.
• bash tests/run once: same single red module.
• python3 bin/perry-lint: 0 errors (3 pre-existing NS-01 warnings).
• git diff --check clean. git diff -- perry/ is empty apart from the dispatched spec.