Skip to content

T5-P1-A2-WP1 — Ensure open_kitchen Failure Predicate Keys on JSON Field#4066

Merged
Trecek merged 1 commit into
developfrom
t5-p1-a2-wp1-ensure-the-open-kitchen-failure-predicate-keys/3999
Jun 11, 2026
Merged

T5-P1-A2-WP1 — Ensure open_kitchen Failure Predicate Keys on JSON Field#4066
Trecek merged 1 commit into
developfrom
t5-p1-a2-wp1-ensure-the-open-kitchen-failure-predicate-keys/3999

Conversation

@Trecek

@Trecek Trecek commented Jun 11, 2026

Copy link
Copy Markdown
Collaborator

Summary

Replace the --- INGREDIENTS TABLE --- substring check in the open_kitchen failure predicate with an "ingredients_table": null JSON field check. This change affects two prompt builders (_build_orchestrator_prompt in cli/_prompts_orchestrator.py and _build_food_truck_prompt in fleet/_prompts.py) and their corresponding test files. The substring marker is injected by a hook and is absent on Codex, causing silent failure-detection blindness on that backend. The JSON field is present in all open_kitchen responses regardless of backend.

Closes #3999

Implementation Plan

Plan file: /home/talon/projects/autoskillit-runs/impl-20260611-073832-079933/.autoskillit/temp/make-plan/T5-P1-A2-WP1-ensure-open-kitchen-predicate-json-field_plan_2026-06-11_074300.md

🤖 Generated with Claude Code via AutoSkillit

Token Usage Summary

Step Model count uncached output cache_read peak_ctx turns cache_write time
plan* opus[1m] 1 1.8k 8.4k 1.1M 80.4k 38 59.5k 7m 3s
verify* sonnet 1 86 8.8k 494.2k 65.2k 32 44.3k 3m 25s
implement* MiniMax-M3 1 679.6k 3.6k 0 0 28 0 2m 22s
audit_impl* sonnet 1 44 4.5k 164.1k 39.7k 11 22.8k 2m 33s
prepare_pr* MiniMax-M3 1 149.7k 1.6k 0 0 12 0 40s
compose_pr* MiniMax-M3 1 175.6k 1.0k 0 0 12 0 37s
review_pr* sonnet 3 366 30.3k 2.0M 58.0k 106 105.7k 8m 31s
Total 1.0M 58.2k 3.8M 80.4k 232.4k 25m 12s

* Step used a non-Anthropic provider; caching behavior may differ.

Token Efficiency

Step LoC Changed cache_read/LoC cache_write/LoC output/LoC
plan 0
verify 0
implement 33 0.0 0.0 109.8
audit_impl 0
prepare_pr 0
compose_pr 0
review_pr 0
Total 33 114629.8 7041.6 1764.6

Model Usage Breakdown

Model steps uncached output cache_read cache_write time
opus[1m] 1 1.8k 8.4k 1.1M 59.5k 7m 3s
sonnet 3 496 43.6k 2.6M 172.8k 14m 30s
MiniMax-M3 3 1.0M 6.3k 0 0 3m 39s

…ailure predicates

The substring '--- INGREDIENTS TABLE ---' is injected by a hook formatter on
Claude Code and is absent on Codex, causing silent failure-detection blindness
on the Codex backend. The JSON field 'ingredients_table' is present in all
open_kitchen responses regardless of backend.

Updated both the orchestrator prompt (cli/_prompts_orchestrator.py) and the
food truck prompt (fleet/_prompts.py) failure predicates to use the JSON
field check. Tests in test_orchestrator_prompt_contract.py renamed and
inverted to assert the new behavior; new test added to
test_food_truck_prompt.py for the food truck prompt.

Refs: T5-P1-A2-WP1

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

@Trecek Trecek left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

AutoSkillit PR Review — Verdict: approved

@Trecek Trecek left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

AutoSkillit review passed. No blocking issues found.

@Trecek Trecek added this pull request to the merge queue Jun 11, 2026
Merged via the queue into develop with commit bae7155 Jun 11, 2026
3 checks passed
@Trecek Trecek deleted the t5-p1-a2-wp1-ensure-the-open-kitchen-failure-predicate-keys/3999 branch June 11, 2026 15:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant