Skip to content

Implementation Plan: T5-P5-A12-WP1 Make diagram staleness enforcement automated#4103

Merged
Trecek merged 4 commits into
developfrom
t5-p5-a12-wp1-make-diagram-staleness-enforcement-automated-a/4031
Jun 13, 2026
Merged

Implementation Plan: T5-P5-A12-WP1 Make diagram staleness enforcement automated#4103
Trecek merged 4 commits into
developfrom
t5-p5-a12-wp1-make-diagram-staleness-enforcement-automated-a/4031

Conversation

@Trecek

@Trecek Trecek commented Jun 13, 2026

Copy link
Copy Markdown
Collaborator

Summary

Create tests/arch/test_recipe_diagram_freshness.py — a parametrized arch test that runs check_diagram_staleness against every bundled recipe YAML. Recipes with pre-rendered diagrams must not be stale (PASS); recipes without diagrams are marked xfail(strict=True) so CI blocks if a diagram is added without removing the xfail marker (XPASS → FAIL). Three meta-tests enforce minimum recipe count (>= 10), xfail cap shrink enforcement (<= 10), and collection sanity. Update tests/arch/AGENTS.md file table for the pre-commit hook invariant.

Implementation Plan

Plan file: /home/talon/projects/autoskillit-runs/impl-20260612-235015-900129/.autoskillit/temp/make-plan/t5-p5-a12-wp1-diagram-freshness_plan_2026-06-12_235500.md

Closes #4031

🤖 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 59 10.9k 778.3k 74.3k 29 78.0k 10m 14s
verify* sonnet 1 4.0k 6.8k 457.2k 49.6k 30 28.6k 3m 38s
implement* MiniMax-M3 1 1.1M 5.0k 0 0 50 0 4m 18s
fix* sonnet 1 190 11.7k 1.1M 58.9k 64 38.5k 5m 56s
audit_impl* sonnet 1 44 7.7k 166.4k 42.0k 12 30.0k 3m 46s
prepare_pr* MiniMax-M3 1 186.4k 2.2k 0 0 14 0 54s
compose_pr* MiniMax-M3 1 175.5k 999 0 0 12 0 35s
review_pr* sonnet 3 398 83.9k 2.4M 80.7k 129 166.3k 18m 46s
resolve_review* sonnet 1 269 20.6k 2.3M 88.9k 78 69.2k 8m 12s
Total 1.4M 149.7k 7.2M 88.9k 410.5k 56m 23s

* 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 77 0.0 0.0 65.6
fix 8 134024.2 4811.9 1456.6
audit_impl 0
prepare_pr 0
compose_pr 0
review_pr 0
resolve_review 14 165901.1 4939.4 1472.6
Total 99 72784.2 4146.5 1512.5

Model Usage Breakdown

Model steps uncached output cache_read cache_write time
opus[1m] 1 59 10.9k 778.3k 78.0k 10m 14s
sonnet 5 5.0k 130.6k 6.4M 332.5k 40m 22s
MiniMax-M3 3 1.4M 8.2k 0 0 5m 46s

autoskillit and others added 4 commits June 13, 2026 00:07
Parametrized arch test that runs check_diagram_staleness against every
bundled recipe YAML. Recipes with pre-rendered diagrams must not be stale
(PASS); recipes without diagrams are marked xfail(strict=True) so CI
blocks if a diagram is added without removing the xfail marker (XPASS
becomes a FAILURE under strict mode).

Meta-tests enforce minimum recipe count (>= 10) and xfail cap shrink
enforcement (<= 10) so the missing-diagram set can only shrink over time.
Updates tests/arch/AGENTS.md file table to keep the sub-AGENTS.md
completeness pre-commit hook passing.

Refs T5-P5-A12-WP1
The diagram freshness test (added in the previous commit) caught four
diagrams with outdated or placeholder hashes: implementation-groups,
merge-prs, and remediation had zero-hash placeholders; implementation
had a hash that predated a subsequent YAML edit. Update each header
comment to match the current compute_recipe_hash() value so
check_diagram_staleness() reports not-stale.
…ts missing diagrams as stale

xfail(strict=True) on recipes without diagrams relies on check_diagram_staleness
returning True for missing files. Add an explicit test to protect this contract
against regression if the function's behavior for absent diagrams is ever changed.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…WITHOUT_DIAGRAMS from set difference

Eliminates the inline set(_RECIPES_WITH_DIAGRAMS) construction inside the
complement comprehension and the separate _RECIPES_WITHOUT_DIAGRAMS intermediate.
_WITHOUT_SET is now computed directly as a set difference, and
_RECIPES_WITHOUT_DIAGRAMS derives from it via sorted(), making the derivation
chain explicit and reducing redundant intermediate state.

Co-Authored-By: Claude Sonnet 4.6 <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 review passed. No blocking issues found.

@Trecek Trecek added this pull request to the merge queue Jun 13, 2026
Merged via the queue into develop with commit 9faa94d Jun 13, 2026
3 checks passed
@Trecek Trecek deleted the t5-p5-a12-wp1-make-diagram-staleness-enforcement-automated-a/4031 branch June 13, 2026 07:58
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