Skip to content

Implementation Plan: Discipline Delivery Channel Matrix Contract#4119

Merged
Trecek merged 3 commits into
developfrom
t5-p6-a1-wp1-encode-the-session-type-x-backend-discipline-de/4044-2
Jun 27, 2026
Merged

Implementation Plan: Discipline Delivery Channel Matrix Contract#4119
Trecek merged 3 commits into
developfrom
t5-p6-a1-wp1-encode-the-session-type-x-backend-discipline-de/4044-2

Conversation

@Trecek

@Trecek Trecek commented Jun 27, 2026

Copy link
Copy Markdown
Collaborator

Summary

Create tests/contracts/test_discipline_delivery_matrix.py — a parametrized 8-case contract test suite that encodes the session-type × backend discipline-delivery channel matrix. The test exercises build_interactive_cmd, build_skill_session_cmd, build_food_truck_cmd, and _read_full_sous_chef composition paths to assert that the PRIMARY delivery channel is populated for each (SessionType × backend) combination, making future regressions CI-detectable. Also update tests/contracts/AGENTS.md with the new file entry.

Closes #4044

Implementation Plan

Plan file: /home/talon/projects/autoskillit-runs/impl-20260626-224909-325309/.autoskillit/temp/make-plan/t5_p6_a1_wp1_discipline_delivery_matrix_plan_2026-06-26_230000.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 2.5k 25.1k 2.3M 130.1k 57 135.7k 16m 5s
verify* sonnet 1 1.5k 5.1k 152.4k 47.8k 15 50.8k 6m 53s
implement* MiniMax-M3 1 69.2k 10.8k 1.2M 0 50 0 4m 36s
audit_impl* sonnet 1 52 12.6k 199.8k 42.0k 16 37.8k 3m 38s
prepare_pr* MiniMax-M3 1 41.7k 3.1k 192.0k 0 13 0 47s
compose_pr* MiniMax-M3 1 36.4k 1.7k 140.9k 0 12 0 32s
review_pr* sonnet 1 134 19.5k 801.9k 67.7k 35 49.8k 5m 41s
resolve_review* sonnet 1 199 22.4k 1.7M 88.9k 55 70.9k 8m 43s
Total 151.7k 100.3k 6.7M 130.1k 345.1k 46m 57s

* 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 178 6686.2 0.0 60.7
audit_impl 0
prepare_pr 0
compose_pr 0
review_pr 0
resolve_review 19 87154.5 3731.4 1178.2
Total 197 33891.5 1751.6 509.0

Model Usage Breakdown

Model steps uncached output cache_read cache_write time
opus[1m] 1 2.5k 25.1k 2.3M 135.7k 16m 5s
sonnet 4 1.9k 59.6k 2.8M 209.3k 24m 56s
MiniMax-M3 3 147.2k 15.6k 1.5M 0 5m 55s

Trecek and others added 3 commits June 26, 2026 23:16
Adds tests/contracts/test_discipline_delivery_matrix.py encoding the
session-type x backend discipline-delivery matrix as a parametrized 8-case
contract suite. Each combination asserts the PRIMARY delivery channel is
populated via the correct builder (--append-system-prompt for Claude,
developer_instructions= for Codex) and that SESSION_TYPE is correctly
threaded through env for headless variants.

Sous-chef composition is verified to flow into orchestrator and open-kitchen
prompts but NOT fleet dispatch.

Updates tests/contracts/AGENTS.md file registry.
…t test

Removes the hardcoded _SESSION_TYPE_VARIANTS=4 integer, the DELIVERY_CHANNEL_CASES
list that used freeform string literals instead of SESSION_TYPE_* constants, and the
self-referential test_delivery_channel_cases_count meta-test whose DELIVERY_CHANNEL_CASES
source was never consumed by any @pytest.mark.parametrize decorator. Adding a backend
to BACKEND_REGISTRY would have failed this count test while leaving the parametrize
lists in each class silently stale, creating false coverage signal. Also removes the
now-unused BACKEND_REGISTRY import.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ractive check

build_interactive_cmd never includes the literal "exec" token — only headless
codex_exec_base commands do. The assertion was always trivially true and could not
detect regressions where the wrong builder variant was used. The preceding
developer_instructions= check is sufficient to verify the interactive primary channel.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@Trecek Trecek added this pull request to the merge queue Jun 27, 2026
@github-merge-queue github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Jun 27, 2026
@Trecek Trecek added this pull request to the merge queue Jun 27, 2026
Merged via the queue into develop with commit c063618 Jun 27, 2026
3 checks passed
@Trecek Trecek deleted the t5-p6-a1-wp1-encode-the-session-type-x-backend-discipline-de/4044-2 branch June 27, 2026 06:52
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