Skills in your scenarios

A skill vendor's results statement shows its executor behaves on the bus's fixtures. Your scenarios show the whole chain works in your house: your configured instance fires, your story owner turns the flag into a gate, your tools hold, your standards desk clears it. Each link is a different product, so each is a place it can break.

Add skill steps to the scenario runs in step 4

Available now, and to rehearsals of the standards-gate and AI-skill

workflows Available now, where the skill executor's stand-in plays until your vendor's executor is cast in its role. The stand-in's warnings are scripted, not raised from your configured instances, so they don't test your values: to test those, cast your vendor's executor.

S4 · The standards gate, step by step

A skill flags a story; the owner gates it; graphics must wait for standards. The run sheet:

#	Actor	Action	Expected on the bus	Check
1	Newsroom system	Create a story that meets your configured instance's condition	story.context seq 1 → 202	Every receiver shows it
2	Skill executor	(automatic)	One skill.warning.raised, flag, rule_id = your label, causation_id = seq 1's message_id	Exactly one warning, within your agreed time. detail reads sensibly to an editor
3	Newsroom system	(automatic, or the editor)	story.context seq 2 with an editorial gate PENDING, naming the skill and your label	The owner turned the flag into a gate (P-13). The editor sees why
4	Graphics	Book the lower third	link.committed, gate status PENDING → 202	Booked, not aired
5	Graphics	Try to clear it itself	link.gate_changed → 403 producer.message_type_not_allowed	Refused: graphics wasn't granted it (step 3)
6	Playout	Try to air it	Nothing	Playout holds while the gate is PENDING
7	Standards desk	Clear it	link.gate_changed CLEARED, then system.audit CLEARED → 202	Recorded, with who cleared it
8	Newsroom system	(automatic, or the editor)	story.context seq 3, the gate approved	The owner closed the gate
9	Playout	Air it	telling.started, then telling.ended → 202	Aired only after step 8
10	Skill executor	(automatic, on seq 2 and seq 3)	Nothing new, unless the warning itself changed	No repeat warnings for the same condition

Passes when graphics and playout never output the item while the gate is PENDING, every reaction traces to its cause, and there's exactly one warning. Step 3 is the one that most often fails: if nobody turns the flag into a gate, the chain stops there and graphics goes ahead unchecked. SOM doesn't say which system does it, so your map must.

Skill cases worth adding

One or two lines each, on top of S1 to S8:

Case	Play	Expect
The condition never holds	A story that doesn't meet any configured instance's condition	No warning. Most evaluations raise nothing
A hold	A story that meets a hold skill's condition	The warning is hold, non_overridable, lists what's withheld in blocks; every tool withholds those fields
Condition clears, then returns	Change the story so the condition stops, then holds again	Nothing when it stops (a warning is never withdrawn on the bus); one new warning when it returns
Killed story	Kill the story while a warning stands	No new warnings on a killed story
Redelivery	Stop and restart the executor mid-run	Still one warning per condition: no duplicates
Narrowed by skills_config	The owner leaves the skill out of the story's active_skills	That configured instance doesn't run on this story (Configuring your house's skills)
Your values changed	Change a value, then run the same story again with a fresh story_id	The warning follows the new values
Two executors, one skill	Two vendors run the same configured instance	Compare their warnings: same condition, same severity, same moment

What to check on every warning

Open it on the Story timeline and read the raw JSON:

• rule_id is your label; skill_id and skill_version are the skill you configured.
• correlation_id is the story's, and causation_id is the snapshot that caused it.
• severity is what your values say; detail tells an editor what was read and what clears it.
• It was accepted. A refused warning never reaches anyone: for your own executor, look under Activity; for a vendor's, ask it for the rule. The fields and their rules are in A warning that validates.

Record the results against each executor in your skill coverage table and your scorecard.