The tools_fails_2 diagnosis (glm-5.3 dropping trailing required fields):
- probe-phase quiz/explain renamed to probe/recall — a shared name carried
the probe call shape into teaching (the first teach quiz dropped nodeId);
display seams accept the old names forever
- teach quiz+explain schemas lead with the required pair (type, nodeId);
explain gains the predict/check type and joins the reveal law (checks
collapse, predictions arrive open; no-type records render check-style)
- plan_init/probe_done results go context-slim (owner, context diet):
content = validated nodeIds + binding-id reminder / fixed ack; details
keeps the tree for the popup and {ok:true} — serializer-verified that
details never reaches the model; failure echoes dropped
- the cards carry the tool's own name (Probe/Recall) and the quiz type
reads on the label (Prediction/Check) in composer and transcript
The whitehead incident (2026-10-09): glm-5.3 emitted quiz type with
quotes baked in ('"predict"') — rejected ten times, byte-identical
retries, the lesson never started — and correctAnswer with the JSON
key-value colon glued on ([':parmenides']). Third GLM wobble incident,
and each prior heal was wired only where its incident landed; this one
landed precisely where the practice heal never reached.
Layer 1 — healEncodedArguments moves to core/argument-heal (the display
seam needs it too — the tool-parts law) with withArgumentHeal +
healToolRecord (agent/tools/heal) wired at EVERY registry: prober,
teacher, practice, bareQuestionTools (it had no prepareArguments at
all), demo-builder, and submit-agent (composing after
healStringifiedArrays; the auditor's extraTools wrapped).
Layer 2 — stripModelValueDebris (core/value-debris): leading/trailing
colons, quotes, backticks, whitespace strip at the MEMBERSHIP seams —
coerceCorrectAnswer (boundary check + display + grading in one),
normalizeQuizType (persisted wobbled records keep their role),
resolveTargetId (copy-syntax on targets).
Layer 3 — schema description hygiene: the quiz type and practice
mode/outcome/rung/missClass values now read as bare words — the quoted
JSON-key style invited quote-copying.
Display mirror — parseQuizInput heals the raw record on read, so a
boundary-healed quiz whose record keeps raw args renders instead of
becoming an invisible pending quiz (learner lockout).
Pins: tests/tool-heal.test.ts (the incident shapes end-to-end through
pi-ai validation, the wiring, the strips, the mirror). The stringy-
options pins flip to boundary-parity per tool-parts' own law; the
chat-flow garbage pin keeps its honest-reject intent with an unhealable
shape. 1471 tests, typecheck, lint green.
The teaching cycle gains its motivation question. The teach quiz schema
now carries a REQUIRED type field — predict (posed before the teaching,
arrives open, the learner predicts the next step) or check (posed after,
arrives collapsed behind the reveal). The prober's quiz stays role-less.
- core: quizInputObject.type union + normalizeQuizType + the reveal law
as pure quizNeedsReveal; QuestionData.questionType; note → thinking
rename (legacy reads fold back); readingThinking channel on answers;
compactAnswerForModel emits thinking / thinkingWhileReading
- tools: descriptions stripped to mechanics — QUIZ_OPTION_CRAFT and
EXPLAIN_DOCTRINE deleted, doctrine moved to the prompt layer
(reverses the 2026-09-27 placement, owner 2026-10-03); shared
quizDescription/explainDescription builders differ only in deltas
- prompts: shared/20-question-tools.md rewritten (option craft,
transcribed-speech leniency, the two thinking channels); new
teacher/28-predictions.md (cycle, roles, skip rule, resolution law);
20-teaching + 30-question-principles scoped to the new taxonomy
- UI: predictions render open with a Prediction eyebrow, verdict shows
but the explanation tucks behind a toggle; the reading-thinking field
mounts while a check waits (dictation claims it until reveal);
answered card separates thinking vs thinking-while-reading blocks
Sycophancy mitigation as a structural commitment: the model writes its
full correctAnswer (plus explanation, in that order) BEFORE the learner
sees the question; the commitment rides the persisted call args, is
revealed on submit, and grades his free text in the model's next reply —
the client never grades free recall.
- tools.explain setting (segmented 'Quizzing — Multiple choice | Free
recall' in Agent settings' Tools group): free arms explain and drops
quiz for BOTH agents (prober variant carries no nodeId, same pattern
as the probe quiz); mechanics keys never pass filterEnabledTools —
quiz stays unwritable (the 2026-09-19 wedge ruling stands)
- doctrine lives pi-natively in the tool descriptions (owner ruling —
no conditional system-prompt chunks): explain carries the commitment
rule; the MC option-building craft moved out of the prompt into the
quiz description; shared/20-question-tools.md slims to the
quiz-agnostic core
- display seams: parseExplainInput (LaTeX-debris restoration included),
dock text-mode with an 'I don't know' row (kbd row 0), AnsweredCard
reveal (The answer / Why, no verdict chips), [Explain answer] format
- phases/plan seams recognize both quizzing tools' nodeId tags; explain
answers deliberately feed no ledger evidence (no client grading to
key on — a future ruling can parse the model's grade)
1049/1049 tests, typecheck/lint/build green.
- src/agent/lh-tool.ts deleted: the zod-wrapped LhTool + toPiTool bridge
(z.toJSONSchema per call, clientQuestion flag, hand-rolled result mapping)
is gone. Tools are pi-agent-core native AgentTools now.
- New src/agent/tool-execute.ts owns what the wrapper added: the abort guard
(runWithAbortGuard: run signal + 30s per-call timeout + 250ms grace,
'Operation aborted' / timeout wording preserved), ToolInputError (the
retryable '- problem at path' format), result mapping (string verbatim /
object JSON / image parts, details rides the record's tool part output),
and questionResult() — placeholder + terminate:true (owner Option A:
belt of terminate + slim makeQuestionStopper with the clean-question
rule; maxTurns deleted from the teaching turn entirely per owner).
- All 11 tool files + tools.ts registries converted to TypeBox parameters
(typebox@1.3.7, pinned to pi-agent-core's copy). zod refinements
(quizInputProblems, askOptionProblems, live-nodeId) moved from schemas
into execute — the model sees the same retryable error shape.
- question-core + contracts/probe TypeBox-authored (single source shared
with the UI); tool-parts parse seam runs the boundary's full chain
(Convert → Check → Clean → refinement), fixing the live-smoke lockout:
pi's boundary Value.Converts correctAnswer: 17 → '17' but the record
keeps raw args, so a parse without Convert rejected a posed quiz
(invisible pending question, turn ended). Pinned with the real smoke
emission shape.
- Subagents converted: demo-builder/researcher AgentTool-native,
visual-maker/pdf-reader declaration-only pi-ai Tools with local execute
fns, extraction submit schemas re-authored in TypeBox (owner option 2:
zod defaults → Type.Optional + fill at the submit seam).
- Accepted delta, flagged: TypeBox-native schemas skip pi's JSON-Schema
primitive coercion fast path... (boundary still Value.Converts), and
prober-quiz unknown keys ride through (Value.Clean strips at display).
- Tests: 988 passing (was 967 at the recon baseline); tool-abort and
read-tool rewritten against the native seams; question-core/question-
tools/tool-parts/demos/web-search/graph-tools/plan-tree/probe-done/
tool-toggles converted to Value.Check/execute pins.
Live smoke (PORT=3004, real model): native ls turn streams end-to-end; a
real quiz pose (first attempt missing correctAnswer → tool-input-error →
model retry → posed) terminates the turn cleanly and persists pending.
The architectural answer to the validation-crash question: validation IS
first at the model-facing boundary (pi validates arguments before
execute; failures return to the model same-turn — all three live
incidents retried successfully). But the same bytes also exist UNVALIDATED
by design in two other places: the live wire streams raw tool-input
deltas DURING generation, and the transcript keeps the raw input of
REJECTED calls (honesty + the replay invariant). The consumers were
supposed to trust part STATES — but that trust was enforced by
TypeScript casts ('part.input as X'), which are compile-time lies with
no runtime check behind them. That is the hole both crashes came
through, and it should not have been possible.
The structural rule now: model-authored INPUT is consumed only through
parsers running the SAME zod schema the model faced — boundary-valid ⇔
consumer-parseable, mechanically. Server/learner-authored OUTPUT stays
trusted (our code authors it).
- src/core/tool-parts.ts (new): parseQuizInput (display parse over the
shared quizInputBase — strips the teacher-only nodeId, accepts both
prober and teacher boundary shapes), parseAskInput, parseProbeDoneInput.
- contracts/probe.ts: probeDoneInputSchema is now core-owned zod — the
probe_done tool declares it (single source of truth), the seam runs
the literal same schema.
- question-dock.tsx: the dock parses its input at entry; unparsable
renders the quiet malformed-question fallback, never a crash.
- probe-done-card.tsx: the input preview goes through the parse; the
old state reasoning and shape guards collapse into it.
Pinned in tests/tool-parts.test.ts: both exact live crash shapes
(stringified quiz options, stringified assessment), boundary-parity
cases for all three schemas, nodeId stripping, mid-stream partials;
render-cards pins the dock fallback markup.
Gates: tsc, lint, build, 328/328.
amosblomqvist