The guard checked whether the installed SDK satisfied the pin. Nobody had
ever checked whether anyone had READ it. Those are different questions, and
the gap between them was a whole version range: pinned >=0.2.111,<0.3,
premises source-verified through 0.2.110, installed 0.2.120. Every build in
between was admissible and unexamined — `uv sync --upgrade` would have kept
806 tests green on an SDK no one had opened. Written red first: a guard
handed 0.2.140 returned it without complaint.
_VERIFIED_THROUGH is the ratchet. It records the newest build actually read
at source, and a newer one fails naming the five premises to re-check. The
pin is untouched and was never the defect — measurement dissolved the
premise that it needed lifting. It was not too narrow but too wide, and a
wider permission is not repaired by widening it further.
The premises themselves were prose the failure message recited. Nothing
tested them, so one that stopped being true would have surfaced on the one
live paid run (S10, D6). They are now a table introspected against the
installed package, with the printed prose derived from that same table so a
checked attribute cannot go unreported or a reported one unchecked. The
premise introspection structurally cannot see — that query() yields an
AssistantMessage then a closing ResultMessage — is named apart, and is the
honest reason the human reading still has to happen.
Value-proved, not merely named: disabling the ratchet reds 1 test, stubbing
the inventory to "no gaps" reds 3, re-hardcoding the prose reds 1, and
lowering _VERIFIED_THROUGH below the installed build reds the real
installed-version test rather than only a monkeypatched one.
0.2.139 read at source (0.2.120 -> 0.2.139, latest on PyPI today; STATE said
0.2.134, measured 08-09 and stale). The public query.py is byte-identical,
every premise field keeps its type and default, and the parser changes are
additive. One needed a look: 0.2.139 added a skills path defaulting
setting_sources to ["user", "project"], which would have undone the S10
isolation fix — it fires only on None, so the explicit [] is out of reach.
Prose carrying stale version claims moved with the reading, never ahead of
it: each was re-verified at 0.2.139 before being restated.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014dKDjVG7qrBh9NkAAxutqN
Operator decision 2026-08-09: the two implementations are developed in the open
against each other. Reading the sibling is permitted and exchange may go both ways;
code is copied only where it genuinely serves the solution, never as a shortcut. The
constraint that remains is what the Microsoft Agent Framework and the Claude Agent
SDK each actually offer — which is the difference the comparison exists to measure.
The old rule lived in three places and two of them would have started lying:
- CLAUDE.md carried the prohibition, and it is read at every session start — left
alone it would have made the next session refuse exactly what was authorised.
- README stated in the present tense that this repo is "built from the shared spec
alone" and "deliberately does not reverse-engineer the MAF sibling". True for
every line written so far, and untrue for anything written from today.
So the README claim is DATED rather than deleted. Through 2026-08-09 the
independence is real and stays claimed; from 2026-08-09 the open-competition rule
applies and is stated plainly, with the consequence spelled out: anything comparing
the two implementations as INDEPENDENT evidence must be scoped to work on or before
that date. The honesty rule (method spec §1) does not let a claim outrun what was
actually done, and that applies to the README's claim about itself.
CHANGELOG is deliberately UNTOUCHED. It describes what was true at that release, and
it was true — 73 commits through 3e91072 were built from the spec alone. A changelog
records history; it is not rewritten because a forward rule changed.
What did NOT change, and is restated in the README so the dating cannot be misread as
a loosening of the method: the spec is still normative and the golden suite is still
the validator's only oracle.
787 tests green, ruff clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JYWMfPKmJtv7JvYnpomkdV
Both links named the private ktg/ namespace. commons published the repo at
open/ (identical history, same HEAD) and reported the four hits it measured
in our working copy. These are the two we own; the two in shared/README.md
are commons-owned and arrive via subtree pull, never edited in place.
README.md:64 is the one that mattered: it sits on the published surface and
promised a clonable source a stranger gets a 404 on.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qr6TwWrHDHeukHy3bL4hgb
Claude Agent SDK verified against official docs + PyPI 2026-07-03 (0.2.110, CLI
bundled, offline import without API key). Contracts mirror method-spec §10/§4.1/§8:
data-source, model-map (per-profile default required), termination (positive caps),
binary feedback decision. TDD: tests written red-first; suite 14/14 green without
any API key; ruff + mypy --strict clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AaQCFnfsh3tfq1VfzdJpoi
