learn(workflow, fixture-seeding): a jump has one fixed target; probed build errors; workflow seed reset (WF-1, WF-5)

[MendixMau]

Direct lane

What changed and why: two workflow text fixes from a five-app lab field report (mxcli v0.26.0, Mendix 11.15). WF-1: workflow-structure-rules.md §1 now says a Jump to has one fixed target and cannot return to the step that sent it. A first workflow design sent seven calls to one shared failure node that "returns" to the caller; that cannot be built, the Stage 3 count caught it (9 breaches), and the redesign cost a resolution round plus a resync of four documents. The rule says: for one failure handler serving N steps, use a child workflow reached through Call workflow, end the child normally and treat failure as data, and know that a parent jump aborts a running child. learned-workflow-patterns.md §9 gains three build-error rows probed with native mx check (CE0117 on a System.Workflow argument, CE1613 at Call workflow, and returns before exposed as workflow action). WF-5: a short "workflow data" block in fixture-seeding.md: abort instances through the runtime first; if SQL is unavoidable, delete the link tables first; reset a context object only after its instances are gone. The §9 rows are filed upstream: mendixlabs/mxcli#1529 (CE0117), #1530 (CE1613, and syntax workflow call-workflow shows the failing form), #1531 (describe drops every call workflow mapping), #1533 (returns parse messages, v0.26.0 "expecting BEGIN" and v0.27.0 "expecting STRING_LITERAL"). WF-5 table names were read from a Mendix 11.15 database: system$workflow_currentactivity, system$workflow_workflowdefinition, system$workflowusertask_assignees, _targetusers, _targetgroups, plus the system$workflowendedusertask_* links. The reusable call-and-wait pattern is a separate branch and is not in this PR.

Impact

Row	Answer
Process	none: no new step, gate or artifact; situational skill text
Dependencies	skills/workflow-structure-rules.md, skills/learned-workflow-patterns.md, skills/fixture-seeding.md, CHANGELOG.md. All three are ondemand in bin/lib/skill-routing.tsv. No new artifact read; this head already contains the earlier workflow-notes change from master.
Context cost	none: net zero always-loaded words. bin/render-routing.sh --check, origin/master and this head: baseline 75225 words (budget 80000) on both, and all nine stage lines (P to 7, every per-role count) are byte-identical, so no trim was needed.
Behaviour	NOT RUN: situational skill text (A/B recommended, not required); the three error rows were probed with native mx check, the jump rule is reasoning from the Mendix docs on Call workflow
Runtime	n/a: no script changed

Field evidence: a five-app lab project (Mendix 11.15, mxcli v0.26.0): one Stage 3 round lost to the shared-failure-node design, one seed reset failing on workflow link tables.

• No client data anywhere in the diff
• Size cap respected (4 files)
• Test tier: T0. Text only, so no test fails on master; render-routing --check passes on both (above) and bin/check-pr-discipline.sh origin/master is clean.
• CHANGELOG lines added under ## Unreleased, both credited

🤖 Generated with Claude Code