fix: sync commandless chat trigger schema

[byapparov]

Refs aictrl-dev/aictrl#3763
Unblocks aictrl-dev/aictrl#4820

Problem

The bundled workflow schema predates commandless chat triggers. The upcoming
aictrl release therefore fails its workflow-schema bundle drift check.

Behavior

• Keeps exact slash-command chat triggers valid without a chat allowlist.
• Allows commandless triggers only with a non-empty chats allowlist.
• Documents both authoring modes and updates schema provenance and checksums.

Scope

• The public schema is validation-equivalent to source commit
  748107cc025bfaab890a4e37161c8e29a64c3c20 after description annotations are removed.
• No runtime, deployment, or workflow-execution code changes.

Rollout and risks

Merge this public bundle before advancing the matching baseline in aictrl.
The aictrl baseline update is already prepared and verified locally.

Verification

• Normalized source and public schemas both hash to 322cc471d1e0a5d9380bd368638710d3e2f694466f9548a719a067f52a038a6e.
• Command, commandless, missing-allowlist, and empty-allowlist contract cases passed.
• All three bundled workflow examples passed schema and DAG validation.
• npm test — 27 tests passed.
• npm run validate — 14 public skills and the public plugin validated.
• Checksum manifest and git diff --check passed.

[aictrl-dev]

🟡 Eval log says "catch-all"; PR says "commandless".

--- a/evals/results.md
+++ b/evals/results.md
@@ -159,7 +159,7 @@ Pending (human): run the scenario in a fresh agent session against a public
 repository pull request and verify that the generated draft grounds every claim
 in the selected commit.
 
-## create-workflow chat catch-all schema sync — 2026-08-22
+## create-workflow commandless chat trigger schema sync — 2026-08-22
 
 Method: synchronized the bundled v1 trigger definitions and authoring guidance
 with source commit `748107cc025bfaab890a4e37161c8e29a64c3c20`, then ran the

🤖 Fix with your agent

Fix this code review finding (aictrl-dev/skills PR #28, evals/results.md:162):

Problem: Eval log says "catch-all"; PR says "commandless"
Detail: The new eval section is titled "create-workflow chat catch-all schema sync", but the schema descriptions, authoring-guide.md, and the PR title all call this feature "commandless". The term "catch-all" is also semantically misleading: commandless mode is the opposite of catch-all — it fires only for chats in a required non-empty allowlist.
Suggested fix: Rename the heading to "create-workflow commandless chat trigger schema sync — 2026-08-22" so the log entry uses the same terminology as the schema and authoring guide it claims to validate.

Suggested patch:
--- a/evals/results.md
+++ b/evals/results.md
@@ -159,7 +159,7 @@ Pending (human): run the scenario in a fresh agent session against a public
 repository pull request and verify that the generated draft grounds every claim
 in the selected commit.
 
-## create-workflow chat catch-all schema sync — 2026-08-22
+## create-workflow commandless chat trigger schema sync — 2026-08-22
 
 Method: synchronized the bundled v1 trigger definitions and authoring guidance
 with source commit `748107cc025bfaab890a4e37161c8e29a64c3c20`, then ran the


Implement the fix on the PR head branch and add a regression test that fails before the fix and passes after.

Why this matters

The new eval section is titled "create-workflow chat catch-all schema sync", but the schema descriptions, authoring-guide.md, and the PR title all call this feature "commandless". The term "catch-all" is also semantically misleading: commandless mode is the opposite of catch-all — it fires only for chats in a required non-empty allowlist.

Pending (human): run the scenario in a fresh agent session against a public
repository pull request and verify that the generated draft grounds every claim
in the selected commit.

## create-workflow chat catch-all schema sync — 2026-08-22

Method: synchronized the bundled v1 trigger definitions and authoring guidance
with source commit `748107cc025bfaab890a4e37161c8e29a64c3c20`, then ran the

[aictrl-dev]

🟡 Empty chats allowlist passes when command present.

🤖 Fix with your agent

Fix this code review finding (aictrl-dev/skills PR #28, skills/create-workflow/reference/v1/workflow.schema.json:236-241):

Problem: Empty chats allowlist passes when command present
Detail: The new allOf/then branch enforces minItems:1 on chats only for commandless triggers; a command trigger with chats: [] still validates. An empty allowlist semantically means the trigger can match no chat (docs say "omit it to fire for any chat", not "set it empty"), yielding a silently dead trigger that passes schema validation. The gap predates this PR (chats never had minItems), but this PR adds the minItems machinery and leaves the empty-array case unconstrained in the command branch. Note: the bundle mirrors the upstream source schema (aictrl-dev/aictrl/schemas/workflow/v1), so the durable fix belongs upstream, then re-synced here.
Suggested fix: If the empty-allowlist semantics are unintended, add "minItems": 1 to the base chats property (next to maxItems: 100) in the source schema aictrl-dev/aictrl/schemas/workflow/v1/workflow.schema.json, so chats: [] is rejected in both branches (the then-branch then only needs required: ["chats"]), and re-sync the bundle. If empty-means-any is intended, document it in the chats description instead.

Implement the fix on the PR head branch and add a regression test that fails before the fix and passes after.

Why this matters

The new allOf/then branch enforces minItems:1 on chats only for commandless triggers; a command trigger with chats: [] still validates. An empty allowlist semantically means the trigger can match no chat (docs say "omit it to fire for any chat", not "set it empty"), yielding a silently dead trigger that passes schema validation. The gap predates this PR (chats never had minItems), but this PR adds the minItems machinery and leaves the empty-array case unconstrained in the command branch. Note: the bundle mirrors the upstream source schema (aictrl-dev/aictrl/schemas/workflow/v1), so the durable fix belongs upstream, then re-synced here.

        "chats": {
          "type": "array",
          "description": "Optional allowlist of chat/channel ids this trigger fires for. Ints (Telegram) or strings (Slack/Discord) are stored as text. Command triggers may omit it to fire for any chat; commandless triggers require at least one entry. Max 100 entries.",
          "items": { "type": ["string", "integer"] },
          "maxItems": 100
        },

[aictrl-dev]

Code review

Verdict: Looks good — only minor / nit comments below. · 🔴 0 · 🟠 0 · 🟡 2 · ⚪ 0 · 0/2 resolved

• 🟡 evals/results.md:162 — Eval log says "catch-all"; PR says "commandless"
• 🟡 skills/create-workflow/reference/v1/workflow.schema.json:236-241 — Empty chats allowlist passes when command present

🤖 Fix all 2 open findings with your agent

Fix the following code review findings on aictrl-dev/skills PR #28 (head branch).
Run the relevant tests/linters after each change.

1. evals/results.md:162 — Eval log says "catch-all"; PR says "commandless"
   Detail: The new eval section is titled "create-workflow chat catch-all schema sync", but the schema descriptions, authoring-guide.md, and the PR title all call this feature "commandless". The term "catch-all" is also semantically misleading: commandless mode is the opposite of catch-all — it fires only for chats in a required non-empty allowlist.
   Suggested fix: Rename the heading to "create-workflow commandless chat trigger schema sync — 2026-08-22" so the log entry uses the same terminology as the schema and authoring guide it claims to validate.
2. skills/create-workflow/reference/v1/workflow.schema.json:236-241 — Empty chats allowlist passes when command present
   Detail: The new allOf/then branch enforces minItems:1 on chats only for commandless triggers; a command trigger with chats: [] still validates. An empty allowlist semantically means the trigger can match no chat (docs say "omit it to fire for any chat", not "set it empty"), yielding a silently dead trigger that passes schema validation. The gap predates this PR (chats never had minItems), but this PR adds the minItems machinery and leaves the empty-array case unconstrained in the command branch. Note: the bundle mirrors the upstream source schema (aictrl-dev/aictrl/schemas/workflow/v1), so the durable fix belongs upstream, then re-synced here.
   Suggested fix: If the empty-allowlist semantics are unintended, add "minItems": 1 to the base chats property (next to maxItems: 100) in the source schema aictrl-dev/aictrl/schemas/workflow/v1/workflow.schema.json, so chats: [] is rejected in both branches (the then-branch then only needs required: ["chats"]), and re-sync the bundle. If empty-means-any is intended, document it in the chats description instead.

📋 Out-of-diff findings (2)

Sev	Location	Finding
🟡	evals/results.md:162	Eval log says "catch-all"; PR says "commandless"
🟡	skills/create-workflow/reference/v1/workflow.schema.json:236-241	Empty chats allowlist passes when command present

Reviewed 5 files · 0 inline · view all 2 findings ↗

---

aictrl · AI code review for fast-moving teams · aictrl.dev

[byapparov]

Finding	Verdict / action	Resolution	Verification
Eval log says catch-all	TRUE · FIX	Renamed the eval heading to use the public contract term commandless chat trigger in c6b8677.	npm test — 27/27 PASS; npm run validate — PASS; checksum manifest and git diff --check — PASS.
Empty chats on a command trigger	FALSE · IGNORE	The claimed dead-trigger impact is contradicted by the upstream contract: chatAllowed intentionally treats absent or empty allowlists as any chat for command triggers, and chat-trigger-adapter-unit.test.ts asserts that an exact command fires with chats: []. Changing only the public mirror would also break its required validation equivalence with source commit 748107cc0.	Upstream runtime test fires for any chat when the allowlist is empty; normalized public/source hash remains 322cc471….

Please re-review exact head c6b8677 for this bounded remediation round.

[aictrl-dev]

Code review

Verdict: Looks good — only minor / nit comments below. · 🔴 0 · 🟠 0 · 🟡 2 · ⚪ 0 · 0/2 resolved

• 🟡 skills/create-workflow/reference/authoring-guide.md:34 — Guide says up to 10 triggers; schema caps at 1
• 🟡 skills/create-workflow/SPEC.md:24-25 — SPEC bundle tree omits reference/v1 schema

🤖 Fix all 2 open findings with your agent

Fix the following code review findings on aictrl-dev/skills PR #28 (head branch).
Run the relevant tests/linters after each change.

1. skills/create-workflow/reference/authoring-guide.md:34 — Guide says up to 10 triggers; schema caps at 1
   Detail: authoring-guide.md (top-level fields comment ~line 34 "up to 10 file-declared event triggers", the Triggers section prose, and the authoring checklist "At most 10 entries in triggers:") claims up to 10 triggers, but the shipped schema that validate.mjs compiles (reference/workflow.schema.json) sets triggers maxItems: 1 with description "at most one trigger per workflow". Pre-existing and not introduced by this PR, but it sits in the same Triggers documentation this PR refreshes for the chat-message contract, leaving a contradicting count uncorrected.
   Suggested fix: Align the guide with the shipped schema: state "at most 1 trigger per workflow" in the top-level fields comment, the Triggers intro, and the authoring checklist (or confirm upstream intent and relax the schema instead).
2. skills/create-workflow/SPEC.md:24-25 — SPEC bundle tree omits reference/v1 schema
   Detail: The Bundle layout tree in SPEC.md lists reference/ as only authoring-guide.md, workflow.schema.json, and examples/, omitting reference/v1/workflow.schema.json — the very artifact this PR re-syncs. The same document's sync procedure and the refreshed Provenance table both depend on that file, and validate.mjs unconditionally reads it (ajv.addSchema of reference/v1/workflow.schema.json), so the inventory contradicts the rest of the SPEC. Pre-existing gap, but this PR edits SPEC.md to update the v1 provenance row without completing the tree.
   Suggested fix: Add a `v1/workflow.schema.json` entry under `reference/` in the Bundle layout tree so the inventory matches the sync procedure, the Provenance table, and validate.mjs's expectations.

📋 Out-of-diff findings (2)

Sev	Location	Finding
🟡	skills/create-workflow/reference/authoring-guide.md:34	Guide says up to 10 triggers; schema caps at 1
🟡	skills/create-workflow/SPEC.md:24-25	SPEC bundle tree omits reference/v1 schema

Reviewed 5 files · 0 inline · view all 2 findings ↗

---

aictrl · AI code review for fast-moving teams · aictrl.dev

[byapparov]

Finding	Verdict / action	Resolution	Verification
Guide says 10 triggers while bundled v2 caps at 1	TRUE · DEFER	The contradiction is real, but changing the guide to one trigger would preserve the wrong side of the drift: upstream v1/runtime support multiple triggers and aictrl#3739 explicitly requires coexisting chat and GitHub triggers. The durable fix is upstream v2 schema parity, followed by a public v2 re-sync; this release-blocking v1 mirror PR does not change that separate schema.	Public v2 maxItems: 1; upstream v1 maxItems: 10; runtime/store coverage supports N triggers.
SPEC tree omits bundled v1 schema	TRUE · FIX	Added reference/v1/workflow.schema.json to the bundle inventory in ccf10b2 and regenerated its checksum.	npm test — 27/27 PASS; npm run validate — PASS; checksum manifest and git diff --check — PASS.

This is the second and final bounded remediation head: ccf10b2.