run_gh_api(paginate=True) breaks on >1 page: JSONDecodeError from concatenated pages, --slurp incompatible with jq=

[cbeaulieu-gt]

Summary

Flagged by CodeRabbit on PR #39 (inline comment, scripts/gh-create-issue-context.py:87), declined as out-of-scope for that PR because the bug lives in pre-existing shared code, not anything #39 introduced.

run_gh_api(paginate=True) in scripts/_gh_common.py passes gh api --paginate and parses stdout with a single json.loads() call. gh api --paginate (without --slurp) emits one JSON value per page, concatenated — so any response spanning more than one page raises JSONDecodeError: Extra data.

CodeRabbit's proposed fix: use --slurp to wrap all pages into a single outer array, then flatten. Caveat: --slurp is mutually exclusive with --jq — any paginate=True call site that also passes jq= needs that jq filtering moved to post-processing.

Affected call sites

run_gh_api(paginate=True) is used identically (no --slurp) by:

• scripts/gh-create-issue-context.py
• scripts/gh-summary.py
• scripts/gh-quick-wins.py
• scripts/gh-refresh-issues.py
• scripts/gh-release-status.py

All are exposed to this bug whenever a paginated query actually spans more than one page.

Suggested fix

• Add --slurp in run_gh_api when paginate=True.
• Flatten the resulting array-of-page-arrays before returning.
• Audit call sites for paginate=True combined with jq= and move that filtering to Python-side post-processing.
• Add a two-page regression test (this is what exposed the gap — CodeRabbit's repro used a JSON round-trip probe, not a live multi-page fetch).

🤖 Generated by Claude Code on behalf of @cbeaulieu-gt