Processes — folio-assistant-core
The diagrams folio-assistant-core holds in its declared processes directories — one sub-graph’s slice of every instance’s processes. The skills, lanes and findings below count these diagrams only.
Generated by cat-harness-tools/scripts/gen-processes-viz.ts. Every number here is counted from the directory, never carried in prose — bpmn-processes’ own rule, and the bean that asked for this page had two different counts in its title and its body.
30 diagram(s), 30 parsed, 0 that would not load.
| where | diagrams |
|---|---|
folio-assistant-core/collab/processes/content/ | 14 |
folio-assistant-core/processes/library/ | 6 |
folio-assistant-core/processes/content/ | 4 |
folio-assistant-core/processes/ui/ | 3 |
folio-assistant-core/collab/processes/conduct/ | 1 |
folio-assistant-core/collab/processes/library/ | 1 |
folio-assistant-core/dsa/processes/ | 1 |
Every process — one page each
Each page shows the diagram, what it is for, who acts in it, every step with the skill it runs, and which processes call it. kg:audit’s process-diagram-published fails a diagram no page shows — 45 of 62 were in that state before these pages existed (bean ooq3).
Running instances links each workflow instance’s own page — the diagram with its token, its stuck states, its beans and todos — for every instance running this process or calling it as a subprocess. Links and counts only: where an instance IS lives in its file under beans/workflows/.
What runs this skill?
The reverse of <bootstrap.processes:skill ref>, which is the join nothing else exposes: a skill says what to do, and until now nothing said which processes ask for it.
61 distinct skill(s) are named by an activity.
Names an ⓘ task panel shows without a link are listed on Task panel links: QA.
Who appears in a process?
A lane NAMES the Role its tasks are performed in; it is not itself a Role. The NAME is free text — process-model.ts notes that sixty lanes spell two dozen positions — so the lane’s partitionElementRef (or the legacy <bootstrap.processes:role ref>) is the join that does not depend on spelling, and it is reported separately below rather than merged into the name.
58 distinct lane name(s).
| lane | in |
|---|---|
| Ingestion Engine (agent, runs unattended) | 6 |
| CI/CD Pipeline | 3 |
| Editor | 3 |
| Review Committee | 3 |
| Authoring agent | 2 |
| Authoring agent (system) | 2 |
| Authoring Committee | 2 |
| Board renderer | 2 |
| Build pipeline | 2 |
| Build pipeline (system) | 2 |
| Editors + authoring agents | 2 |
| Folio-Assistant Agent | 2 |
| Public review (commenters, review committee, editor) | 2 |
| Publication manager | 2 |
| QA reviewer (independent) | 2 |
| Reader | 2 |
| Adaptation committee (policy, clinical, programme) | 1 |
| Adaptation team (human or agentic) | 1 |
| Author | 1 |
| Author (human or agentic) | 1 |
| Author (person) | 1 |
| Author / guideline developer (person) | 1 |
| Author and policy author | 1 |
| Change set — feature branch and staging | 1 |
| Clinical / scientific SMEs | 1 |
| Content Author | 1 |
| Deep researcher | 1 |
| Editor / author (person) | 1 |
| Evidence agent (system) | 1 |
| Guideline publisher | 1 |
| HCI validation pipeline (system) | 1 |
| Intake agent | 1 |
| Layout layer | 1 |
| Log watcher (authoring agent) | 1 |
| Mechanical validation (system) | 1 |
| Non-mechanical validation (review agent or SME) | 1 |
| Onboarding agent (system) | 1 |
| Parties and signatories (provider, recipient) | 1 |
| Programme manager | 1 |
| Programme manager (release authority) | 1 |
| Public commenter | 1 |
| Publish — GitHub Pages | 1 |
| Publisher (data provider’s publication manager) | 1 |
| Re-pin steward and build pipeline | 1 |
| Release build and archive (pipeline) | 1 |
| Review committee | 1 |
| Review committee and editor | 1 |
| Review coordinator | 1 |
| Review Coordinator | 1 |
| Review team (content · QC · technical officer) | 1 |
| Review team and SMEs | 1 |
| Reviewer | 1 |
| Reviewer (legal, policy) | 1 |
| Reviewer (SME or editor) | 1 |
| Trusted registries and observatories (external) | 1 |
| User (person) | 1 |
| Validation and QA (mechanical + agents) | 1 |
| Work plan (beans) | 1 |
Which steps touch the work plan?
<cat-harness.processes:bean op> — the audit trail for who reserves and resolves work.
| op | diagrams |
|---|---|
note | 10 |
Strict, advisory — and nobody said
Three states, not two. loadProcessModel reads an undeclared policy as strict, which is right for the engine — “a process that forgot to say is governed, not exempt” — and hides a fact a reader wants. Somebody chose strict and nobody said, so the engine assumed strict are different, and the second is 9 of 30 here.
| enforcement | declared | defaulted |
|---|---|---|
strict | 9 | 9 |
advisory | 12 | 0 |
Findings
Every one of the 30 diagrams has a rendered SVG.
Activities carrying no <bootstrap.processes:skill ref> — a census, not a gap list
14 across 4 diagram(s). This section reported them as defects in its first version, and that was wrong: beans luke and uuhu worked this corpus from 90 down to this remainder and settled what is left. luke: “coverage is deliberately NOT gated — a human sign-off step has no skill”. uuhu added the call-activity exemption and recorded that its own remainder “are not gaps”.
3 are call activities, which delegate to a subprocess — the skill is named there, and naming it twice would be one fact in two places. The other 11 are listed by lane, because the lane is what says whether a person, a pipeline or an agent performs the step, and only the last of those has a skill to run.
| lane | steps |
|---|---|
| Parties and signatories (provider, recipient) | 6 |
| Author and policy author | 2 |
| Content Author | 1 |
| Editor / author (person) | 1 |
| Folio-Assistant Agent | 1 |
No verdict is offered on which of these is a gap, and that is the honest state rather than a hedge: distinguishing a person’s judgement step from an agent step somebody forgot needs the lane’s actor KIND, which a free-text lane name does not give — process-model.ts notes sixty lanes spell two dozen positions. The lane’s partitionElementRef is the join that would answer it, and it is present on 11 of these 11 steps — which is the measurement, not an impression. An earlier draft of this sentence said few of these and was counting whether the lane had a NAME, a different question with a different answer.