Summary

Fixes #311.

rtk gh api with --jq was returning type placeholders (int, string, { tag_name: string }) instead of real values.

Root cause

run_api() unconditionally passed all gh api output through json_cmd::filter_json_string(), which is a schema extractor — it replaces every JSON value with its type descriptor. When --jq is present, gh has already applied the user's filter and returned targeted output (e.g. a scalar 104 or a small object). Running that through the schema extractor destroys the data the user asked for.

Fix

Detect --jq/--jq=<expr> in args and print the output as-is. Adds a has_jq_flag() helper following the existing should_passthrough_run_view() pattern.

Before / After

# Before
rtk gh api search/issues --method GET -f q='repo:rtk-ai/rtk is:issue' -f per_page=1 --jq '.total_count'
# → int   ❌

rtk gh api "/repos/rtk-ai/rtk/releases?per_page=1" --jq '.[0] | {tag_name,published_at}'
# → { published_at: string, tag_name: string }   ❌

# After
rtk gh api search/issues --method GET -f q='repo:rtk-ai/rtk is:issue' -f per_page=1 --jq '.total_count'
# → 104   ✅

rtk gh api "/repos/rtk-ai/rtk/releases?per_page=1" --jq '.[0] | {tag_name,published_at}'
# → {"published_at":"2026-02-28T21:10:20Z","tag_name":"v0.23.0"}   ✅

Tests

5 new unit tests added to gh_cmd::tests:

• test_has_jq_flag_long_form — --jq .filter style
• test_has_jq_flag_equals_form — --jq=.filter style
• test_has_jq_flag_absent — no flag present
• test_has_jq_flag_empty — empty args
• test_api_jq_output_not_schema_filtered — demonstrates filter_json_string("104") returns "int", confirming why the passthrough is necessary
• test_api_jq_object_would_be_schema_mangled — demonstrates object projections lose real values without the fix

All 28 gh_cmd tests pass.