A warning that validates

An executor publishes exactly one kind of message on a skill's behalf: skill.warning.raised. Build it from the SOM 1.0 schema, not from the sample warnings in the skill files.

The skill library's samples don't validate

The 0.2.2 samples use a wrn-… warning_id (the schema wants a UUID), a null skill_warning_ref (it wants a string) and, in one skill, a system_type outside SOM's list. SOM 1.0 refuses all three, and so does the bus. Load Skill warning copied from the skill library sample in the Playground to see the verdict. RND's position: P-01.

The envelope

Field	Value	Source
som_version	"1.0.0"	SOM
message_id	A UUID (UUIDv7 recommended), new per warning, and the same on a retry of that warning	SOM, P-10
correlation_id	The trigger's correlation_id: the story's	SOM
causation_id	The trigger's message_id. Always set it, although SOM makes it optional	Skill library
message_type	"skill.warning.raised"	SOM
timestamp	When the evaluation produced the warning, RFC 3339	SOM
originating_system	A system_id your connection may stamp; a system_type from SOM's list	SOM, the bus, P-02
topic	"som.skill.warning.raised"	SOM (must start som.)

system_type must be one of SOM 1.0's values. The library's own tool types (rundown, mam, playout, cms…) aren't, apart from graphics and archive, and are refused with envelope.enum. Which value to use is P-02.

The payload

All twelve required fields, and nothing the schema doesn't define.

Field	Rule	Common mistake
warning_id	A UUID (UUIDv7), new for each warning	"wrn-7f2c9a41", copied from the samples
skill_id	The upstream id, smart-stories/<name>, never renamed	–
skill_version	The version of the skill file you implement: "0.2.2" in this suite	"v2"
story_id	The snapshot's story_id	–
scope	story:<story_id>, asset:<asset_id> or link:<link_id>	An instance reference. SOM says never, although the schema only checks that it's a string
severity	hold, flag or inform, lower case, within the skill's range	"FLAG", "warning"
rule_id	The configured instance's label	–
non_overridable	true for hold, otherwise false	Left out
affected_fields	Field paths in the story, such as ["lifecycle.phase"]	–
detail	Plain words: what was read, what it was compared with, and what clears it	–
blocks	[] unless severity is hold; then what is withheld	Left out when empty
skill_warning_ref	A string naming the declaration	null, copied from the samples. See P-09
message_type	Optional. If present, "skill.warning.raised"	A different value: refused with payload.message_type_mismatch

A complete warning

raise-flag-on-match, configured instance house-breaking-indicative-category, triggered by snapshot 6 of the hurricane run:

{
  "som_version": "1.0.0",
  "message_id": "0199a1c4-7b90-7d12-9c3e-5a1f0b2c3d41",
  "correlation_id": "0199a1c4-5b10-7c3a-9e4f-2a6b8c0d1e22",
  "causation_id": "0199a1c4-7a2e-7b31-8c55-4d2f9e6a1b07",
  "message_type": "skill.warning.raised",
  "timestamp": "2026-09-11T16:41:00.412000Z",
  "originating_system": { "system_id": "planning-01", "system_type": "ncs" },
  "topic": "som.skill.warning.raised",
  "payload": {
    "warning_id": "0199a1c4-7b90-7d12-9c3e-5a1f0b2c3d42",
    "skill_id": "smart-stories/raise-flag-on-match",
    "skill_version": "0.2.2",
    "story_id": "hurricane-2026-0911",
    "scope": "story:hurricane-2026-0911",
    "severity": "flag",
    "rule_id": "house-breaking-indicative-category",
    "non_overridable": false,
    "affected_fields": ["lifecycle.phase"],
    "detail": "indicative-category declared: lifecycle.phase reads BREAKING, which matches BREAKING. Cleared by an assertion at or above duty-editor on this story.",
    "blocks": [],
    "skill_warning_ref": "house-breaking-indicative-category/story:hurricane-2026-0911/1",
    "message_type": "skill.warning.raised"
  }
}

Checkpoint

• [ ] My warnings pass the Playground: warning_id a UUID, skill_warning_ref a string, system_type from SOM's list.
• [ ] Every warning carries the trigger's correlation_id, and its message_id as causation_id.

Next: How an executor behaves.