Fix Scheduled schedule block: time_cycle/tz, not cron/timezone

The Scheduled-trigger eval failed at release even after adding event: Schedule — 'missing timer_event_definition or schedule parameters for trigger'. Root cause (proven live end-to-end: import + release + delete): the schedule block's fields are time_cycle and tz, NOT cron and timezone. The wrong names import cleanly but fail release. Grounded in real Foundry sample workflows (anomali-threatstream, insider-risk-sailpoint/workday, ngsiem-importer), which all use schedule: {time_cycle, start_date, end_date, tz, skip_concurrent}.

Scanning all foundry-sample-* workflows also surfaced a second valid shape: an event: Schedule trigger with NO schedule block at all (rapid-response, scalable-rtr) — a caller-scheduled job template where cadence is supplied at execution time. The validator must not reject that.

Changes:
- validate.py: flag schedule blocks using cron:/timezone: with rename hints to time_cycle:/tz:, and flag a schedule block missing time_cycle. Only fires when a schedule block is present, so caller-scheduled templates (no block) pass.
- both trigger-types.md copies: correct the schedule block field names and document the caller-scheduled-template variant.
- tests: wrong-field-names flagged, correct fields pass, no-block template passes.

Verified live: a workflow with schedule: {time_cycle, tz, ...} imports AND releases (definition since deleted); the cron/timezone shape imports but fails release.

Author: mraible

skills/authoring/references/trigger-types.md

@@ -254,22 +254,33 @@ Runs on a cron-like schedule.
 
 Like a Signal trigger, a Scheduled trigger MUST carry an `event:` field naming its
 category, which is `Schedule`. Omitting it fails import with
-`code 2003: "unknown trigger event named "` — even when the `schedule:` block is
-present and well-formed. The cron and timezone live in `schedule:`; the `event:`
-is a separate, required top-level field.
+`code 2003: "unknown trigger event named "`. The schedule itself goes in a
+`schedule:` block whose fields are **`time_cycle`** (the cron expression) and
+**`tz`** (the timezone) — NOT `cron:`/`timezone:`. Those wrong names import but
+fail at release with "missing timer_event_definition or schedule parameters for
+trigger". Include `start_date`/`end_date` (empty strings unless bounding the
+window) and `skip_concurrent`.
 
 ```yaml
 trigger:
     next:
         - FirstAction
-    name: Scheduled
     event: Schedule            # REQUIRED — the trigger category
     type: Scheduled
     schedule:
-        cron: "0 */6 * * *"    # Every 6 hours
-        timezone: UTC
+        time_cycle: "0 */6 * * *"   # cron expression — the field is time_cycle
+        start_date: ""              # empty unless bounding the start
+        end_date: ""                # empty unless bounding the end
+        tz: Etc/UTC                 # timezone — the field is tz, not timezone
+        skip_concurrent: true       # skip a run if the previous one is still going
 ```
 
+A `schedule:` block is required only for a self-scheduled recurring workflow. An
+`event: Schedule` trigger with **no** `schedule:` block is also valid — that is a
+caller-scheduled job template, where the run cadence is supplied at execution time
+rather than baked into the definition (several Foundry sample workflows ship this
+shape).
+
 **Warning**: Disable scheduled workflows after testing to avoid rate limiting.
 
 ---

skills/authoring/scripts/validate.py

@@ -869,6 +869,37 @@ def structural_check(file_path):
                 "timezone go in the 'schedule:' block; the 'event' is separate."
             )
 
+        # The Scheduled 'schedule' block's cron/timezone fields are named
+        # 'time_cycle' and 'tz' — NOT 'cron'/'timezone'. The wrong names import
+        # cleanly but fail at release with "missing timer_event_definition or
+        # schedule parameters for trigger" (confirmed live). Flag them here so the
+        # error surfaces at authoring time instead of release. Only checked when a
+        # 'schedule:' block is present — an `event: Schedule` trigger with NO
+        # schedule block is a valid on-demand job template (the schedule is
+        # supplied by the caller at runtime), as several Foundry samples show.
+        schedule = trigger.get("schedule")
+        if isinstance(schedule, dict):
+            if "cron" in schedule and "time_cycle" not in schedule:
+                issues.append(
+                    "ERROR: Scheduled trigger's schedule uses 'cron:' — the release "
+                    "field is 'time_cycle:'. 'cron' imports but fails release with "
+                    "'missing timer_event_definition or schedule parameters'. Rename "
+                    "'cron' to 'time_cycle'."
+                )
+            if "timezone" in schedule and "tz" not in schedule:
+                issues.append(
+                    "ERROR: Scheduled trigger's schedule uses 'timezone:' — the "
+                    "release field is 'tz:'. Rename 'timezone' to 'tz' (e.g. "
+                    "'tz: Etc/UTC')."
+                )
+            if "time_cycle" not in schedule:
+                issues.append(
+                    "ERROR: Scheduled trigger's schedule block has no 'time_cycle' "
+                    "(the cron expression, e.g. '0 */6 * * *'). A schedule block "
+                    "must carry it, or omit the block entirely for a caller-scheduled "
+                    "job template. See references/trigger-types.md."
+                )
+
         # Signal triggers fire from an event and carry no caller-supplied input,
         # so a 'parameters' schema on them is invalid: any field it declares
         # becomes an undefined variable at release ("Signal triggers do not

skills/workflows/references/trigger-types.md

@@ -196,22 +196,33 @@ Runs on a cron-like schedule.
 
 Like a Signal trigger, a Scheduled trigger MUST carry an `event:` field naming its
 category, which is `Schedule`. Omitting it fails import with
-`code 2003: "unknown trigger event named "` — even when the `schedule:` block is
-present and well-formed. The cron and timezone live in `schedule:`; the `event:`
-is a separate, required top-level field.
+`code 2003: "unknown trigger event named "`. The schedule itself goes in a
+`schedule:` block whose fields are **`time_cycle`** (the cron expression) and
+**`tz`** (the timezone) — NOT `cron:`/`timezone:`. Those wrong names import but
+fail at release with "missing timer_event_definition or schedule parameters for
+trigger". Include `start_date`/`end_date` (empty strings unless bounding the
+window) and `skip_concurrent`.
 
 ```yaml
 trigger:
     next:
         - FirstAction
-    name: Scheduled
     event: Schedule            # REQUIRED — the trigger category
     type: Scheduled
     schedule:
-        cron: "0 */6 * * *"    # Every 6 hours
-        timezone: UTC
+        time_cycle: "0 */6 * * *"   # cron expression — the field is time_cycle
+        start_date: ""              # empty unless bounding the start
+        end_date: ""                # empty unless bounding the end
+        tz: Etc/UTC                 # timezone — the field is tz, not timezone
+        skip_concurrent: true       # skip a run if the previous one is still going
 ```
 
+A `schedule:` block is required only for a self-scheduled recurring workflow. An
+`event: Schedule` trigger with **no** `schedule:` block is also valid — that is a
+caller-scheduled job template, where the run cadence is supplied at execution time
+rather than baked into the definition (several Foundry sample workflows ship this
+shape).
+
 **Warning**: Disable scheduled workflows after testing to avoid rate limiting.
 
 ---

tests/test_validate.py

@@ -376,6 +376,46 @@ def test_scheduled_trigger_with_event_ok(self, tmp_path):
         issues = validate.structural_check(str(f))
         assert not any("missing an 'event'" in i for i in issues)
 
+    def test_scheduled_schedule_cron_field_flagged(self, tmp_path):
+        # A schedule block using 'cron:'/'timezone:' (instead of time_cycle/tz)
+        # imports but fails release; flag the wrong field names at authoring time.
+        f = tmp_path / "scheduled_wrong_fields.yaml"
+        content = VALID_WORKFLOW.replace(
+            "  type: On demand",
+            "  type: Scheduled\n  event: Schedule\n"
+            "  schedule:\n    cron: '0 8 * * *'\n    timezone: UTC",
+        )
+        f.write_text(content)
+        issues = validate.structural_check(str(f))
+        assert any("'cron:'" in i and "time_cycle" in i for i in issues)
+        assert any("'timezone:'" in i and "tz" in i for i in issues)
+
+    def test_scheduled_schedule_correct_fields_ok(self, tmp_path):
+        # A schedule block with the release-valid field names passes.
+        f = tmp_path / "scheduled_right_fields.yaml"
+        content = VALID_WORKFLOW.replace(
+            "  type: On demand",
+            "  type: Scheduled\n  event: Schedule\n"
+            "  schedule:\n    time_cycle: '0 8 * * *'\n    tz: Etc/UTC\n"
+            "    start_date: ''\n    end_date: ''\n    skip_concurrent: true",
+        )
+        f.write_text(content)
+        issues = validate.structural_check(str(f))
+        assert not any("time_cycle" in i for i in issues)
+        assert not any("schedule uses" in i for i in issues)
+
+    def test_scheduled_no_schedule_block_is_valid_template(self, tmp_path):
+        # 'event: Schedule' with NO schedule block is a valid caller-scheduled job
+        # template (several Foundry samples ship this) — must not be flagged for a
+        # missing time_cycle.
+        f = tmp_path / "scheduled_template.yaml"
+        content = VALID_WORKFLOW.replace(
+            "type: On demand", "type: Scheduled\n  event: Schedule"
+        )
+        f.write_text(content)
+        issues = validate.structural_check(str(f))
+        assert not any("time_cycle" in i for i in issues)
+
     def test_condition_without_default_or_expression_fails(self, tmp_path):
         # A condition with a bare 'next:' and neither 'default: true' nor a
         # cel_expression/expression is what the release-time validator rejects