refactor(ci): switch full-suite cron to twice daily at 01:00/13:00 UTC

Replaces the every-4-hours schedule (00, 04, 08, 12, 16, 20 UTC) with two
daily runs at 01:00 and 13:00 UTC. Singleton concurrency is unchanged, so
only one full-suite run is in flight at a time. Updates the four governance
references that called out the old "every 4 hours" cadence (AGENTS.md, the
ci-debug + smart-test playbooks, and the transcript-change hook summary)
to match.

Signed-off-by: jgstern-agent <josh-agent@iterabloom.com>

Author: jgstern-agent

.agent/agent_playbooks_protocols_sops_skills/ci-debug-protocol.md

@@ -1,3 +1,4 @@
+<!-- SPDX-License-Identifier: AGPL-3.0-or-later -->
 ## CI Debugging Protocol
 When CI fails but tests pass locally, use `./scripts/ci-debug`:
 
@@ -14,7 +15,7 @@ When CI fails but tests pass locally, use `./scripts/ci-debug`:
 
 **CI workflow topology:**
 - **`ci.yml`**: Fast per-PR check (smart-test on changed packages). Gates merge.
-- **`full-suite.yml`**: Periodic validation (every 4 hours + manual dispatch). Runs all packages in parallel. Does NOT trigger on push to dev — with 20+ merges/day and singleton concurrency, the queue never clears. Stop-the-line fires from scheduled runs, so there may be a delay between a breaking merge and the andon cord — this is expected, not a bug.
+- **`full-suite.yml`**: Periodic validation (twice daily at 01:00 and 13:00 UTC + manual dispatch). Runs all packages in parallel. Does NOT trigger on push to dev — with 20+ merges/day and singleton concurrency, the queue never clears. Stop-the-line fires from scheduled runs, so there may be a delay between a breaking merge and the andon cord — this is expected, not a bug.
 - **`nightly.yml`**: Runs at 11 PM UTC. Multi-Python matrix (3.10–3.13) and integration tests. Sets commit statuses (`nightly/test-matrix`, `nightly/integration-tests`) so release.yml can skip them when the release SHA was already covered. `ci-debug status` works for nightly runs too.
 - **`release.yml`**: Triggered by version tag push or manual dispatch. Security audit is a hard gate before publish. Test-matrix and integration-tests are deferred: if nightly already covered the SHA they're skipped, otherwise they run post-publish as verification.
 

.agent/agent_playbooks_protocols_sops_skills/smart-test-playbook.md

@@ -34,7 +34,7 @@ TOTAL                   33003     10    99%
 FAIL Required test coverage of 100% not reached. Total coverage: 99.97%
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 Full output: .ci/pytest-output.log
-Targeted run (14 test files, 3 changed sources). Full suite runs in CI every 4 hours (not post-merge); running it now would be wasteful.
+Targeted run (14 test files, 3 changed sources). Full suite runs in CI twice daily at 01:00 and 13:00 UTC (not post-merge); running it now would be wasteful.
 ```
 
 The Missing column shows exact line numbers to fix. **Do NOT re-run tests** just to find missing lines. The "Targeted run" message means smart-test already validated 100% coverage for your changed files — do NOT follow up with `pytest --full`.

.agent/hooks/_shared/on_transcript_change.py

@@ -268,7 +268,7 @@ def _extract_transcript_metadata(transcript_path: str) -> dict[str, str]:
     ("ci-debug-protocol",
      ".agent/agent_playbooks_protocols_sops_skills/ci-debug-protocol.md",
      "When CI fails but tests pass locally, use ci-debug runs/status/analyze-deps. Four CI "
-     "workflows: ci.yml (per-PR smart-test), full-suite (every 4 hours, all packages), "
+     "workflows: ci.yml (per-PR smart-test), full-suite (twice daily at 01:00 and 13:00 UTC, all packages), "
      "nightly (multi-Python matrix + integration tests), release (on tag push). Common root "
      "causes: missing pyproject.toml dependencies, version mismatches, platform differences. "
      "Never use pytest.skip() to hide failures."),

.ci/affected-tests.txt

@@ -1,11 +1,12 @@
 # Test selection manifest
-# Generated by smart-test at 2026-04-26T19:11:02-04:00
+# Generated by smart-test at 2026-04-26T19:52:54-04:00
 # Mode: targeted
-# Baseline: da884304c90ca449941831de752e6890206d70a1
-# Reason: no Python source files changed
+# Baseline: afcaca6256e7ffc90467fa5544fbd02490edeb47
 # Changed files: 5
 # Changed source files: 0
-# Selected tests: 0
+0
+# Selected tests: 1
 #
 # === CHANGED_SOURCE_FILES ===
 # === SELECTED_TESTS ===
+tests/test_on_transcript_change.py

.github/workflows/full-suite.yml

@@ -8,12 +8,12 @@ name: Full Test Suite
 on:
   # Manual trigger for debugging
   workflow_dispatch:
-  # Periodic validation (every 4 hours)
+  # Periodic validation (twice daily at 01:00 and 13:00 UTC)
   # Post-merge push trigger removed: with 20+ merges/day and singleton
   # concurrency, the queue never clears. Periodic runs catch regressions
   # within hours while saving CI resources.
   schedule:
-    - cron: '0 */4 * * *'  # Every 4 hours
+    - cron: '0 1,13 * * *'  # 01:00 and 13:00 UTC
 
 # Singleton concurrency - only one full suite runs at a time
 # This prevents resource contention and ensures clear pass/fail state

AGENTS.md

@@ -247,7 +247,7 @@ External contributors: see `docs/CONTRIBUTOR_MODE.AGENTS.md` for fork-based work
 Agent runs `prepare-release VERSION` (bumps version, updates changelog, runs release-check, creates dev-to-main PR). Human merges the PR and runs `tag-release VERSION` to create a GPG-signed tag, triggering the release CI workflow. Separation ensures branch protection and human authorization. (For more explanation, please read `.agent/agent_playbooks_protocols_sops_skills/release-workflow.md`.)
 
 ## CI Debugging Protocol
-When CI fails but tests pass locally, use `ci-debug runs/status/analyze-deps`. Four CI workflows: ci.yml (per-PR smart-test), full-suite (every 4 hours), nightly (multi-Python matrix + integration), release (on tag). Common root causes: missing pyproject.toml deps, version mismatches, platform differences. Never poll CI manually. (For more explanation, please read `.agent/agent_playbooks_protocols_sops_skills/ci-debug-protocol.md`.)
+When CI fails but tests pass locally, use `ci-debug runs/status/analyze-deps`. Four CI workflows: ci.yml (per-PR smart-test), full-suite (twice daily at 01:00 and 13:00 UTC), nightly (multi-Python matrix + integration), release (on tag). Common root causes: missing pyproject.toml deps, version mismatches, platform differences. Never poll CI manually. (For more explanation, please read `.agent/agent_playbooks_protocols_sops_skills/ci-debug-protocol.md`.)
 
 ## Testing Optional Dependencies
 For PyPI-available tree-sitter grammars: add to pyproject.toml, write real tests, no mocking. For build-from-source grammars (built via `scripts/build-source-grammars`): write real tests calling the analyzer directly, plus a mock test only for the unavailability code path. Never use `pytest.mark.skipif` as an escape hatch. (For more explanation, please read `.agent/agent_playbooks_protocols_sops_skills/optional-dependency-testing-playbook.md`.)