Repository navigation
feat: model context sources in BPMN - #7
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Task documentation already uses Quelle for provenance; a store's location is a different thing. Spec key quelle becomes ort. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…-flow fixtures Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Six wissen stores, all reading the notebook "The Product - Business Design". Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ries Convention-detected inputs and outputs are candidates; analyze records the one the user picks. Requiring a workflow-* kind for every candidate would mislabel intermediate artifacts as process outputs. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…stores Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The spec gains six contextSources, workflowIO and the store elements. Each of the 19 tasks that read a Product-Methodik store gets its own cited knowledge/<taskId>.md from the notebook "The Product - Business Design". Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ntext sources The five human checkpoints read the knowledge stores too, so 24 tasks have their own cited knowledge file; 11 keep the phase-level mapping. The entry skill asks for feedback, backlog and fachkonzept. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
`argument-hint: [a] [b]` is not valid YAML. Found while generating the user-story-refinement example, whose input has three required fields; the single-field smoke test had passed because `[a]` parses as a list. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The turquoise sat too close to the orchestrator teal in the mapping view. Rose (#F8BBD0 / #880E4F) is a hue no other kind uses and keeps the contrast the script enforces (5.85:1 on its fill, 9.45:1 on white). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A reader of a wissen store has its own knowledge/<taskId>.md. The new install-knowledge.mjs installs each as references/<taskId>.md in the skill that owns the task: footnotes keep the source and drop the verbatim quote and FAQ id, pointers between files are fixed, long files get a contents line, and the owner's Kontextquellen section lists one line per store. Merging the files into domain-knowledge.md would collide on footnote numbers. Found by the quality review of the user-story-refinement example: the FAQ ids cannot be followed at run time and the quotes doubled each file (-42% bytes). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Per-task knowledge files installed in the skills, the entry skill takes feedback, backlog and fachkonzept as input and links the phase files at its gateways, the Gw_Erkenntnis decision is applied, and the mapping view shows the stores in rose. Content fixes from the quality review went into knowledge/ so a regeneration reproduces them. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
BPMN-NOTATION.md lists every BPMN element with what it means when a workflow and its context are modelled for lanecraft and what the pipeline generates from it: roles, task types, control flow, events, phases, data objects, data stores (Art x Ort), process input/output, unsupported constructs, pattern choice and the mapping-view colours. The README links it from "Was jedes Element bedeutet". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
install-knowledge.mjs replaced the whole "## Kontextquellen" section with the wissen lines, so a re-run dropped the live and gedächtnis bullets that generate step 5 writes. It now carries them over under "Live and memory stores, used at run time", and only points at references/domain-knowledge.md when the skill has one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…itHub
The product backlog is no longer two data objects but three live data
stores on the same GitHub backlog (Art: live, Ort: cli:gh), cut by phase.
The workflow writes into it at five places, each right after its own
approval userTask: a note on an open story (I3 -> I4), an Opportunity
Backlog entry (I5 -> I6), the merge into a duplicate (P9 -> P6), and after
"Story priorisieren & Übernahme ins Backlog freigeben" the new story in the
Ready column (P8) and the marks on affected stories (P10). "Betroffene
Stories zur Anpassung markieren" becomes a proposal step ("Anpassungen an
betroffenen Stories vorschlagen").
Re-ran analyze, knowledge, design (two architect reviews), generate and
verify (RESULT: PASS, no red). New: the write skill story-backlog-publish
(approval check, exact approved text via --body-file, marker-based
idempotent resume), a PreToolUse write guard that asks on every non-read
gh call, and settings.json with the gh read permissions. Example README,
README and ARCHITECTURE describe the live backlog.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ran trim-the-fat over bpmn-authoring, bpmn-process-design, bpmn-to-agentic-workflow and the five bpmn2agent-* stages (SKILL.md and references only; scripts, templates, schema and fixtures untouched). 21,799 -> 21,089 words. Headings, step numbers, names and descriptions are unchanged, so cross-references still resolve. Dropped rationale that changes no decision, stated duplicated rules once (tool read/write heuristic in the mapping rubric, context-source routes in verify) and fixed the laneSet position in bpmn-authoring to match its element order. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
skill-template.md, agent-template.md and mapping-report-template.md point to "mapping-rubric.md's language rule", which did not exist. The rubric now states it: BPMN labels verbatim in meta.language, everything else English, except the generated README and the fixed German tokens. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds examples/lanecraft/ with the pipeline modelled in its own notation: six lanes, seven human checkpoints, both pipeline loops with caps and nine data stores of all three kinds. Validated and inventory-clean; not yet run through the pipeline. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Worum es geht
Kontext ist am Ende Daten. Ab jetzt modelliert das Diagramm nicht nur Übergaben (Data Objects), sondern auch woher jeder Schritt sein Wissen holt und wohin er schreibt: als Data Store mit
Art:(wissen|live|gedächtnis) undOrt:(notebook:,mcp:,cli:, URL,datei:,websearch). Lesen und Schreiben ergeben sich aus der Pfeilrichtung. Dazu kommen prozessweite Ein- und Ausgaben. Plan und Aufgaben:docs/plans/kontextquellen/.Was die Pipeline daraus macht
contextSourcesundworkflowIOund fragt bei fehlendemArt:/Ort:oder ungültiger Kombination nach.knowledge/<taskId>.md.tools:oder inpermissions.allow.## Kontextquellenin Skills und AgentspermissionDecision: "ask") und Gedächtnis-Cap-Hookargument-hintaus der Prozess-Eingabereferences/userTaskist ein Fehler.Weitere Änderungen
bpmn-authoring: Data Stores und Prozess-I/O zeichnen; Fixturescontext-flow.bpmnundcontext-flow-unguarded.bpmnbpmn-process-design: fragt beim Modellieren nach den KontextquellenBPMN-NOTATION.mdals Referenz der Notation;README.mdundARCHITECTURE.mdbeschreiben die Kontextquellenexamples/user-story-refinement:examples/lanecraft: lanecraft selbst als BPMN-Diagramm/trim-the-fatgestrafft;grilling-Skill hinzugefügtPrüfung
npm run validategrün, mit einer Warnung: DieCLAUDE.mdim Plugin-Root wird nicht als Kontext geladen.examples/dark-factory/tools/dark-factory-gen/regenerate.sh:RESULT: PASS,no reference problems, die drei Smoke-Zeilen undmapping view okexamples/user-story-refinement:RESULT: PASS, mit 9 Data Stores, 2 Prozess-Eingaben und 9 Prozess-Ausgaben. Alle 5 Schreibzugriffe auf Live-Stores sind durch eine Freigabe geschützt.🤖 Generated with Claude Code