Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Add or evolve ContextOS runtime contracts across TypeScript types, JSON Schemas, canonical examples, documentation, and drift tests. Use for RunContext, ContextPack, CompiledContext, ToolEnvelope, DecisionRecord, ReplayPacket, or delegation changes; not for prose-only edits.
.claude/skills/contextosai-contextos-contract-sync/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 108% | 0% |
Keep every executable representation of a runtime contract semantically aligned and versioned. Do not treat successful JSON parsing as proof of compatibility.
Read the repository instructions, then locate:
$defs dependencies;Read references/surface-map.md for the current contract bundle, parity invariants, and test routing. Confirm the active repository still matches that map before editing.
Classify the change:
Separate schema compatibility from runtime trust. An additive field may remain optional for old payloads while the canonical producer always emits it; consumers must treat absence as legacy/unknown, never trusted.
$id conventions.trace_id, tenant, identity, and lineage propagate across request, compiled context, tools, decision, and replay.ActionRisk has identical definitions where declared, compiled, and executed.Run targeted schema-surface, runtime-schema, spec-drift, compiler, and context-admission tests as applicable, followed by typecheck and the full test suite. Parse every published schema and inspect the exact public schema file set.
Return a compatibility note, changed-surface matrix, migration implications, test evidence, and any consumer behavior that remains unverified.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,158 | 6,861 | +11% | 1 | 1 | 0% | 309 | 984 | +218% | 0 | 0 | — |
case-02 | fail→fail | 5,261 | 6,558 | +25% | 1 | 1 | 0% | 317 | 921 | +191% | 0 | 0 | — |
case-03 | fail→fail | 32,599 | 6,569 | -80% | 1 | 1 | 0% | 5,668 | 827 | -85% | 0 | 0 | — |
case-04 | pass→pass | 12,567 | 14,102 | +12% | 1 | 1 | 0% | 2,166 | 2,960 | +37% | 0 | 0 | — |
case-05 | pass→fail | 11,589 | 5,927 | -49% | 1 | 1 | 0% | 1,268 | 875 | -31% | 0 | 0 | — |
case-06 | pass→fail | 8,861 | 4,050 | -54% | 1 | 1 | 0% | 1,137 | 835 | -27% | 0 | 0 | — |
case-07 | fail→pass | 17,812 | 3,547 | -80% | 1 | 1 | 0% | 2,909 | 1,155 | -60% | 0 | 0 | — |
case-08 | fail→pass | 16,178 | 3,431 | -79% | 1 | 1 | 0% | 2,637 | 1,211 | -54% | 0 | 0 | — |
case-09 | pass→pass | 14,838 | 14,111 | -5% | 1 | 1 | 0% | 2,184 | 2,802 | +28% | 0 | 0 | — |
case-10 | fail→pass | 15,834 | 5,600 | -65% | 1 | 1 | 0% | 2,289 | 1,448 | -37% | 0 | 0 | — |
case-11 | pass→pass | 18,274 | 15,415 | -16% | 1 | 1 | 0% | 3,431 | 3,196 | -7% | 0 | 0 | — |
case-12 | fail→pass | 19,048 | 9,023 | -53% | 1 | 1 | 0% | 2,447 | 1,882 | -23% | 0 | 0 | — |
case-13 | pass→pass | 12,092 | 11,422 | -6% | 1 | 1 | 0% | 1,942 | 2,127 | +10% | 0 | 0 | — |
case-14 | pass→pass | 9,973 | 6,119 | -39% | 1 | 1 | 0% | 1,457 | 1,424 | -2% | 0 | 0 | — |
case-15 | fail→pass | 34,179 | 18,618 | -46% | 1 | 1 | 0% | 1,894 | 3,935 | +108% | 0 | 0 | — |
case-16 | pass→pass | 12,932 | 5,374 | -58% | 1 | 1 | 0% | 2,007 | 1,531 | -24% | 0 | 0 | — |
case-17 | pass→pass | 11,087 | 10,073 | -9% | 1 | 1 | 0% | 1,533 | 2,005 | +31% | 0 | 0 | — |
case-18 | fail→pass | 14,381 | 8,740 | -39% | 1 | 1 | 0% | 1,960 | 1,841 | -6% | 0 | 0 | — |
case-19 | pass→pass | 11,843 | 10,718 | -9% | 1 | 1 | 0% | 1,779 | 2,319 | +30% | 0 | 0 | — |
case-20 | pass→pass | 18,767 | 9,639 | -49% | 1 | 1 | 0% | 1,763 | 1,991 | +13% | 0 | 0 | — |
case-21 | pass→pass | 7,297 | 5,935 | -19% | 1 | 1 | 0% | 993 | 1,514 | +52% | 0 | 0 | — |
case-22 | fail→pass | 16,137 | 9,061 | -44% | 1 | 1 | 0% | 2,467 | 2,184 | -11% | 0 | 0 | — |
case-23 | fail→pass | 8,943 | 4,694 | -48% | 1 | 1 | 0% | 1,372 | 1,319 | -4% | 0 | 0 | — |
case-24 | fail→pass | 9,612 | 2,993 | -69% | 1 | 1 | 0% | 1,521 | 939 | -38% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted, and 19 counted toward the lift figure. The other 5 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +29 percentage points is the difference between those two pass rates over the 19 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.