amazon-science/ammo / ai_cli_session/.codex/skills/ammo
ammo
UnverifiedProfile and optimize GPU kernels for vLLM inference on NVIDIA GPUs. Use when targeting a specific model, hardware, dtype, parallelism, and workload to improve latency.
tests/test_state_schema_semantics.py
View on GitHub# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
# SPDX-License-Identifier: CC-BY-NC-4.0
"""Task-2 regressions for the merged Codex AMMO state schema."""
from __future__ import annotations
import json
from pathlib import Path
from jsonschema import Draft202012Validator
SCHEMA_PATH = Path(__file__).resolve().parents[3] / "schemas" / "state.schema.json"
def _schema() -> dict:
return json.loads(SCHEMA_PATH.read_text(encoding="utf-8"))
def _campaign_properties() -> dict:
return _schema()["properties"]["campaign"]["properties"]
def _round_properties() -> dict:
return _campaign_properties()["rounds"]["items"]["properties"]
def _track_properties() -> dict:
return _round_properties()["parallel_tracks"]["properties"]["tracks"]["additionalProperties"]["properties"]
def _audit_gate_properties(stage: str) -> dict:
return _round_properties()["audit"]["properties"][stage]
def test_new_campaign_floor_is_half_percent_without_deflation_contract():
assert "4.2" in _campaign_properties()["schema_version"]["enum"]
config = _campaign_properties()["config"]
assert config["properties"]["min_e2e_improvement_pct"]["default"] == 0.5
assert "lossless_e2e_deflation_factor" not in config["required"]
assert "lossy_quant_e2e_deflation_factor" not in config["required"]
assert "lossless_e2e_deflation_factor" not in config["properties"]
assert "lossy_quant_e2e_deflation_factor" not in config["properties"]
legacy_floor = config["properties"]["debate_elimination_score_floor"]["description"]
assert "LEGACY" in legacy_floor
assert "EV_pct < min_e2e_improvement_pct" in legacy_floor
assert "backward compatibility" in legacy_floor
def test_round_schema_adds_mining_invalidation_and_diluted_rollup():
props = _round_properties()
assert props["mining_invalidated"]["type"] == "boolean"
assert props["mining_invalidated_reason"]["type"] == ["string", "null"]
diluted_tracks = props["diluted_tracks"]
assert diluted_tracks["type"] == ["array", "null"]
assert diluted_tracks["items"]["properties"]["op_id"]["type"] == "string"
def test_track_schema_preserves_codex_fields_and_adds_diluted_marker():
props = _track_properties()
assert props["diluted"]["type"] == ["boolean", "null"]
assert "SKIPPED" in props["gate_5_1a"]["enum"]
assert "SKIPPED" in props["gate_5_2"]["enum"]
for field in (
"gate_5_1a_skip_reason",
"gate_5_2_skip_reason",
"branch",
"worktree_path",
"evidence_path",
"kill_criteria_results",
"completed_at",
"implementer_agent",
"implementer_rollout_id",
"monitor_agent",
"monitor_evidence_path",
"monitor_offsets_path",
"monitor_summary_path",
):
assert field in props
tracks = _round_properties()["parallel_tracks"]["properties"]["tracks"]
assert tracks["propertyNames"]["pattern"] == "^[A-Za-z0-9_-]+$"
boundary = props["gate_5_2_boundary"]
assert "e2e_equivalent_improvement_pct" in boundary["required"]
def test_selected_candidate_preserves_winning_proposal_locator():
selected = _round_properties()["debate"]["properties"]["selected_candidates"]["items"]
assert "proposal_file" in selected["properties"]
assert "rounds/{N}/debate/proposals/" in selected["properties"]["proposal_file"][
"description"
]
assert "proposal_file" in selected["required"]
def test_selected_candidate_carries_optional_display_name():
"""The lead mints a short display name with each op_id; the op_id stays the
key. Optional for back-compat: pre-4.2 selected_candidates omit it."""
selected = _round_properties()["debate"]["properties"]["selected_candidates"]["items"]
name = selected["properties"]["name"]
assert name["type"] == "string"
assert name["minLength"] == 1
assert "name" not in selected["required"]
def test_selected_candidate_supports_typed_contingent_host_spike():
selected = _round_properties()["debate"]["properties"]["selected_candidates"][
"items"
]
assert selected["properties"]["selection_mode"]["enum"] == [
"ordinary",
"contingent_host_spike",
]
obligations = selected["properties"]["stage_4_validation_obligations"]
assert "production_boundary_spike" in obligations["items"]["enum"]
def test_mining_schema_distinguishes_f_e2e_from_f_decode():
mining = _round_properties()["bottleneck_mining"]["properties"]
assert "decode wall time" in mining["top_f_decode_pct"]["description"]
assert "total E2E wall time" in mining["top_f_e2e_pct"]["description"]
assert "Stage 7 stop check" in mining["top_addressable_e2e_pct"]["description"]
def test_shipped_optimization_can_record_diluted():
item = _campaign_properties()["shipped_optimizations"]["items"]
assert item["properties"]["diluted"]["type"] == ["boolean", "null"]
def test_completed_mining_is_not_globally_deadlocking_schema():
mining = _round_properties()["bottleneck_mining"]
assert "allOf" not in mining
def test_auditor_escalation_accepts_track_array_scope():
spec = _campaign_properties()["auditor_escalation"]
payload = {
"stage": "4_5_parallel_tracks",
"round": 2,
"reason": "three audit cycles exhausted",
"verdict_files": ["rounds/2/audits/stage_45.md"],
"scope": ["op001", "op007"],
"resumable_task": "fix and re-audit quarantined tracks",
}
assert list(Draft202012Validator(spec).iter_errors(payload)) == []
def test_auditor_escalation_accepts_campaign_scope_and_rejects_other_string():
spec = _campaign_properties()["auditor_escalation"]
payload = {
"stage": "6_integration",
"round": 3,
"reason": "campaign-wide invariant failed",
"verdict_files": ["rounds/3/audits/stage_67.md"],
"scope": "campaign",
"resumable_task": "resolve the campaign blocker and re-audit",
}
validator = Draft202012Validator(spec)
assert list(validator.iter_errors(payload)) == []
payload["scope"] = "op001"
assert list(validator.iter_errors(payload))
def test_schema_version_enum_is_unchanged_by_the_audit_start_fields():
# started_at/cycle are additive on the existing 4.2 gate objects. A new
# version entry would force every consumer to learn a version it does not
# need.
assert _campaign_properties()["schema_version"]["enum"] == [
"3.0", "4.0", "4.1", "4.2"
]
def test_audit_gates_carry_optional_started_at_and_cycle():
for stage in ("stage_2", "stage_67"):
gate = _audit_gate_properties(stage)
props = gate["properties"]
started = props["started_at"]
assert started["type"] == ["string", "null"]
assert started["format"] == "date-time"
assert "ammo-audit-start-stamp.sh" in started["description"]
cycle = props["cycle"]
assert cycle["type"] == ["integer", "null"]
assert cycle["minimum"] == 1
# Optional: a legacy gate that carries only passed_at still validates
# against the schema. The 4.2 backstop lives in ammo_state.py, not here.
assert "started_at" not in gate.get("required", [])
assert "cycle" not in gate.get("required", [])
def test_audit_gate_accepts_the_stamped_shape_and_rejects_cycle_zero():
spec = _audit_gate_properties("stage_45")
validator = Draft202012Validator(spec)
stamped = {
"started_at": "2026-08-04T12:00:00Z",
"cycle": 2,
"passed_at": "2026-08-04T12:31:00Z",
"verdict_file": "rounds/1/audits/stage_45_cycle_2.md",
}
assert list(validator.iter_errors(stamped)) == []
assert list(validator.iter_errors({"cycle": 0}))
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |