Benchmark: isolate page-intent skill input - #98
Merged
Conversation
noeltock
force-pushed
the
codex/issue-90-page-engine
branch
from
September 11, 2026 12:18
43f0922 to
c4c08f8
Compare
noeltock
force-pushed
the
codex/issue-90-page-engine
branch
from
September 11, 2026 13:09
c4c08f8 to
6d349b3
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
The page-intent benchmark engine loaded the whole skill bundle and tied its cache identity to registered-block authoring sources it never uses. Closes #90.
Solution
The page engine now reads only
GUIDE.mdthenASSEMBLE.md, and hashes that exact ordered text in its prompt identity and provenance. The publicblock-runner skillcommand still reads the complete bundle.Diff
+180 −58 · 6 files · no public API or model call changeTesting & verification
Reviewed revision:
43f0922npm run typecheck— passed.npx --no-install vitest run dev/test/benchmark.test.ts dev/test/skill.test.ts dev/test/engine-skill.test.ts— passed.npx --no-install vitest run dev/test/cli.test.ts -t 'prints the agent guide with skill'— passed.git diff --check— passed.Mocked filesystem/subprocess tests cover exact ordering, Unicode, missing selected input, newline normalisation, and selected versus excluded reference changes. The selected prompt is 27,566 UTF-8 source bytes at this revision; no model-quality, latency, or score claim was measured.
Risk / rollout
Historical records are untouched. The changed identity prevents reuse of the old full-bundle cache; reverting this commit restores the former prompt coupling.