spec: Count runs from disabled selected workflows

[yihanzhu]

A selected workflow can disappear from the quota preflight count after it is disabled. This spec defines using the existing GitHub CLI option to include its runs and a hermetic count/threshold regression while preserving current selection and failure rules.

Tracks #132. G1: #401. Current intake acceptance: #132 (comment)

review_size: standard

Only work/disabled-workflow-run-counting/spec.md changes. Risk is high because the helper is a runaway safety control; a separate accepted plan still precedes implementation.

Validation: exact merged intent hash, sole-path diff, clean checkout and whitespace check verified. Required quick CI and independent G2 review are pending. No implementation or live provider operation is included.

[yihanzhu]

Independent G2 review — PR #402

Reviewer: /root/package_rebinding_intent_author, independent of spec author /root/disabled_workflow_spec_author. Requested model/effort: gpt-6-astra / high; actual runtime identity is not independently exposed.

Repository: yihanzhu/ystack
PR: #402
Stage: G2 specification and risk classification only
Branch: ystack/spec/disabled-workflow-run-counting

Reviewed-head: 4a93cde
Reviewed-base: ca7daca
Spec-blob: c70ccd50c18858f8165da66d5dce64707d1cbbe1
Intent-blob: d9bd79f672384b5b97d43ab96e9165038fb5d3c3
Intake-title-sha256: 8d8797714fd0b602d816d0e6fb9ad3499faa3b10e797104f4486dfbec12629d5
Intake-body-sha256: 628e11bfe6ede8a92d47bef7a7953cbe7b67610d90dfa87c235a23b057012809

Evidence and identity.

I read the entire 72-line specification, accepted intent, PR description, applicable G2 criteria and current policy. I compared the design with the existing quota helper and hermetic test.

Independent API reads before and after review returned the same PR head, base, title/body and intake hashes above. Issue #132 remained open. PR #402 remained open with no labels; merge-ready was absent. Final remote reads matched the reviewed head and base.

Current intake acceptance is comment 5772185718, verified as authored by yihanzhu. G1 PR #401 is merged at ca7dacab5f0a9ee24c7400cbbd3319858982dfd5. The specification’s intent-blob matches that merged intent exactly.

The candidate consists of one single-parent commit directly above the reviewed base. Only work/disabled-workflow-run-counting/spec.md changes. The checkout is clean and the whitespace check passes.

Bugs pass — no Important findings.

The design addresses the demonstrated cause: add the existing --all option to each selected workflow’s gh run list --workflow query. Installed CLI help confirms the option includes disabled workflows. Retaining the workflow filter and other arguments preserves selection, window and fetch-limit behavior.

The existing global sanity check, accumulation and error handling remain unchanged. Requirements preserve numeric validation, fatal listing failures, unexpected per-workflow failures and the existing missing-workflow behavior.

Both intent questions are answered concretely. The existing stub will model disabled-workflow visibility. With one enabled workflow contributing 3 runs, disabled counts of 16 and 17 must produce totals of 19 and 20, with the corresponding success and threshold refusal. The original query would instead report only 3. Requiring the regression to fail with that original query establishes meaningful behavioral proof.

The specification retains every existing assertion and requires the complete affected suite, pinned ShellCheck and required CI on the final implementation. It does not claim those future checks have passed.

Security pass — no Important findings.

The change preserves the runaway brake’s threshold, selection, exclusions, configuration precedence and failure rules. Errors cannot become zero usage.

Prospective implementation is limited to the existing helper and its test. No workflow, permission, credential, executable-selection, quota-policy, activation or target boundary expands. The already implemented run-attempt identity remains intact. Hermetic proof requires no real workflow operation or provider invocation.

Classifying this change as risk: high is appropriate because it changes a safety control. The small patch size does not remove the separately accepted high-risk plan requirement.

Compliance pass — no Important findings.

The spec has the required frontmatter, valid intent hash link, explicit risk, verifiable requirements, bounded design, proof, exclusions and areas of concern. Its plain-language content is sufficient for a separate plan author without relying on conversation history.

The design stays within the accepted G1 outcome and the existing GitHub adapter boundary. It adds no portable-core dependency, exceptional implementation path or new framework. No conflict with the accepted intent or north star was found.

The PR tracks #132 without closing it. The spec identifies the implementation’s standard review size; the PR description records its own standard size. Neither artifact includes a review-round changelog. No implementation or accepted upstream artifact changes are present.

CI and conclusion.

The API reports quick-CI run 35696646815 completed successfully at the reviewed head. This is the observed automatic run, not implementation or full-suite proof. The manager must still verify required checks and unchanged identities before protected merge.

No Important findings or nits. Accept this exact specification and its high-risk classification at G2. A separately reviewed and accepted high-risk plan must land before implementation; this verdict supplies no code or merge authorization.

This reviewer made no file, ref, worktree, label, comment or other state changes and executed no tests or targets.

Review-verdict: ACCEPT