test(evals): add waza eval suite for dx-ticket-analyze#144
Open
easingthemes wants to merge 2 commits into
Open
test(evals): add waza eval suite for dx-ticket-analyze#144easingthemes wants to merge 2 commits into
easingthemes wants to merge 2 commits into
Conversation
Six task scenarios (bare ID, full ADO URL, Jira key, no-arg, with/without project.yaml) with structural + behavioral graders. CI runs mock executor on every PR; live runs gated to workflow_dispatch. https://claude.ai/code/session_01SzqQhaY1BUwTjHXALddv7h
Verified against waza v0.31.0 binary locally. Changes: - Use inline `content:` on inputs.files[] (ResourceRef has no `source` field). This lets each task carry its own .ai/config.yaml variant (ADO vs Jira). - Move shared fixtures (project.yaml, component-index.md) under fixtures/.ai/project/ so paths mirror the workspace layout waza requires. - Switch model to `claude-haiku-4.5` (the short name waza accepts). - CI uses `--skip-graders` for mock runs since mock output is dummy and won't satisfy real graders. Live runs (workflow_dispatch with copilot-sdk) drop the flag and grade for real. Mock CI now exits 0 on a clean run, catching schema/fixture/dispatch regressions while leaving behavior validation to live runs. https://claude.ai/code/session_01SzqQhaY1BUwTjHXALddv7h
easingthemes
pushed a commit
that referenced
this pull request
Jun 16, 2026
Three pillars that turn a static harness into one that compounds with use: 1. Evals — measure quality per skill, regression detection in CI 2. Learning loop — PR review comments + step failures → proposed rule diffs for human approval (never auto-commits) 3. Pattern promotion — cross-ticket clustering → proposed skills and references/ entries Connects existing TODOs #1 #52 #65 #67 #100 #143 #144 into a coherent system design. No new TODOs needed — what was missing was the architecture that shows how they form a flywheel. https://claude.ai/code/session_01YE5UZvuFpEyAhAH24Ds9Sd
easingthemes
pushed a commit
that referenced
this pull request
Jun 16, 2026
Three pillars that turn a static harness into one that compounds with use: 1. Evals — measure quality per skill, regression detection in CI 2. Learning loop — PR review comments + step failures → proposed rule diffs for human approval (never auto-commits) 3. Pattern promotion — cross-ticket clustering → proposed skills and references/ entries Connects existing TODOs #1 #52 #65 #67 #100 #143 #144 into a coherent system design. No new TODOs needed — what was missing was the architecture that shows how they form a flywheel. https://claude.ai/code/session_01YE5UZvuFpEyAhAH24Ds9Sd
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
2 participants
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.
Six task scenarios (bare ID, full ADO URL, Jira key, no-arg, with/without
project.yaml) with structural + behavioral graders. CI runs mock executor
on every PR; live runs gated to workflow_dispatch.
https://claude.ai/code/session_01SzqQhaY1BUwTjHXALddv7h