drawio-databricks

Name: drawio-databricks
Author: sparklabx

111installs104forks

How to install drawio-databricks?

sparklabx/drawio-ai-kit/drawio-databricks

$npx -y skills add sparklabx/drawio-ai-kit --skill drawio-databricks

Installs into the current project.

Summary

Use when the user asks for a Databricks lakehouse architecture diagram — medallion architecture (Bronze/Silver/Gold), Delta Lake, Unity Catalog, workspace deployment, data-plane/control-plane, or any diagram built with Databricks icons. Builds with the declarative layout engine u

Files · 1

View on GitHub

SKILL.md

1	# Draw.io Databricks
2
3	Produce correct Databricks lakehouse architecture diagrams in draw.io. This skill
4	is a thin frontend; the deterministic engine, validator, and rules live in the
5	`drawio-ai-kit` package, reached via the `drawio-ai` CLI.
6
7	## 0. Preflight — the CLI must be installed
8
9	```bash
10	command -v drawio-ai >/dev/null 2>&1 \|\| echo "Install the Kit first: npm i -g github:sparklabx/drawio-ai-kit"
11	```
12
13	If `drawio-ai` is not on PATH, stop and tell the user to run
14	`npm i -g github:sparklabx/drawio-ai-kit`. Never run `npm i -g` yourself — nothing mutates the
15	user's global environment without their say-so.
16
17	## 1. Delegate the build (preferred when your harness supports it)
18
19	If your harness can spawn autonomous subagents that run shell commands AND read
20	images (e.g. Claude Code's Task tool, a general-purpose agent), run the whole
21	build loop in a subagent — the rules, icon searches, and every render/fix
22	iteration then cost this conversation nothing. If it can't (or the subagent
23	can't read images), skip to Inline path below — same loop, same rules.
24
25	Before spawning, resolve what the subagent cannot ask about: diagram scope,
26	output directory (absolute path under the user's project), filename. Run the
27	preflight above yourself. For a multi-diagram request, spawn one subagent per
28	diagram in parallel with distinct filenames.
29
30
31	Model routing — if your harness lets you choose the subagent's model, route by
32	task weight: a fast/cheap tier (Claude Haiku-class — must support vision) when
33	the request matches a template from the rules' Templates table (reproduction is
34	mechanical; the validator's advice strings teach every fix), your **default strong
35	model** for free-hand or novel architectures. If a cheap subagent returns VALIDATE
36	not ok or ITERATIONS > 3, respawn ONCE on the strong model before taking over
37	inline. Multi-diagram requests: route each diagram independently.
38
39	Subagent prompt (fill every `<...>`):
40
41	```text
42	Build a Databricks lakehouse architecture .drawio diagram with the drawio-ai CLI.
43	Request: <user's request + clarifications, verbatim>
44	Output: <ABS_PROJECT_DIR>/<NAME>.drawio — never write inside the Kit, never into cwd.
45	Follow exactly:
46	1. Set ROOT="$(drawio-ai root)". Read $ROOT/docs/api-cheatsheet.md — the full layout-engine
47	API in one file; never read library source.
48	2. Run `drawio-ai workflow` and `drawio-ai principles --mode databricks` — the source of
49	truth. (Fallback if a command is blocked: read $ROOT/rules/*.md directly.)
50	3. Look up every icon with ONE batched `drawio-ai search "a, b, c"`; never recolor icons.
51	4. Scaffold, don't write: `drawio-ai scaffold --list`, pick the closest template, then
52	`drawio-ai scaffold <name>.mjs -o <dir>/build.mjs` — the script arrives runnable and
53	self-checking. Edit only the deltas; Write a new script only if no template is close
54	AND you'd change more than half of it.
55	5. Each `node build.mjs` run prints validate JSON AND the render's machine-readable
56	`issues` list. Fix from THAT checklist — all issues in one Edit round — then re-run.
57	Loop until issues is empty.
58	6. Only when issues is empty: Read the PNG once as final visual confirmation. Target <= 2
59	PNG reads total. Then render once WITHOUT --check for the final deliverable PNG.
60	Do NOT invoke any drawio skill — this prompt already contains the full procedure.
61	Do not ask questions — make the standard choice and record it under ASSUMPTIONS.
62	Return EXACTLY this block, nothing else:
63	DRAWIO: <absolute path to .drawio>
64	PNG: <absolute path to .png>
65	VALIDATE: <verbatim final validate JSON>
66	ICONS: <comma-separated icon names used>
67	ITERATIONS: <number of render/fix cycles>
68	SUMMARY: <one sentence describing the diagram>
69	ASSUMPTIONS: <choices made without asking, or "none">
70	```
71
72	Relay `DRAWIO`, `PNG` and `SUMMARY` to the user verbatim; do NOT re-read the
73	.drawio or PNG in this conversation — the subagent already ran the vision
74	self-check. If `VALIDATE` is not ok, take over via the Inline path (the build
75	.mjs and .drawio are on disk at the returned paths).
76
77	## Inline path (no subagent support)
78
79	### 1. Shared Workflow
80
81	```bash
82	drawio-ai workflow
83	```
84
85	Prints the build → validate → render → write-to-project-path loop every diagram
86	follows. Read it; it is the source of truth for the process.
87
88	### 2. Domain rules
89
90	```bash
91	drawio-ai principles --mode databricks
92	```
93
94	Returns the Databricks rules + shared principles + catalog categories.
95
96	### 3. Build with the engine, then validat

People who liked this also viewed

6 picks

Type

drawio-databricks

111installs104forks

View on GitHub

How to install drawio-databricks?

sparklabx/drawio-ai-kit/drawio-databricks

$npx -y skills add sparklabx/drawio-ai-kit --skill drawio-databricks

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Draw.io Databricks
2
3	Produce correct Databricks lakehouse architecture diagrams in draw.io. This skill
4	is a thin frontend; the deterministic engine, validator, and rules live in the
5	`drawio-ai-kit` package, reached via the `drawio-ai` CLI.
6
7	## 0. Preflight — the CLI must be installed
8
9	```bash
10	command -v drawio-ai >/dev/null 2>&1 \|\| echo "Install the Kit first: npm i -g github:sparklabx/drawio-ai-kit"
11	```
12
13	If `drawio-ai` is not on PATH, stop and tell the user to run
14	`npm i -g github:sparklabx/drawio-ai-kit`. Never run `npm i -g` yourself — nothing mutates the
15	user's global environment without their say-so.
16
17	## 1. Delegate the build (preferred when your harness supports it)
18
19	If your harness can spawn autonomous subagents that run shell commands AND read
20	images (e.g. Claude Code's Task tool, a general-purpose agent), run the whole
21	build loop in a subagent — the rules, icon searches, and every render/fix
22	iteration then cost this conversation nothing. If it can't (or the subagent
23	can't read images), skip to Inline path below — same loop, same rules.
24
25	Before spawning, resolve what the subagent cannot ask about: diagram scope,
26	output directory (absolute path under the user's project), filename. Run the
27	preflight above yourself. For a multi-diagram request, spawn one subagent per
28	diagram in parallel with distinct filenames.
29
30
31	Model routing — if your harness lets you choose the subagent's model, route by
32	task weight: a fast/cheap tier (Claude Haiku-class — must support vision) when
33	the request matches a template from the rules' Templates table (reproduction is
34	mechanical; the validator's advice strings teach every fix), your **default strong
35	model** for free-hand or novel architectures. If a cheap subagent returns VALIDATE
36	not ok or ITERATIONS > 3, respawn ONCE on the strong model before taking over
37	inline. Multi-diagram requests: route each diagram independently.
38
39	Subagent prompt (fill every `<...>`):
40
41	```text
42	Build a Databricks lakehouse architecture .drawio diagram with the drawio-ai CLI.
43	Request: <user's request + clarifications, verbatim>
44	Output: <ABS_PROJECT_DIR>/<NAME>.drawio — never write inside the Kit, never into cwd.
45	Follow exactly:
46	1. Set ROOT="$(drawio-ai root)". Read $ROOT/docs/api-cheatsheet.md — the full layout-engine
47	API in one file; never read library source.
48	2. Run `drawio-ai workflow` and `drawio-ai principles --mode databricks` — the source of
49	truth. (Fallback if a command is blocked: read $ROOT/rules/*.md directly.)
50	3. Look up every icon with ONE batched `drawio-ai search "a, b, c"`; never recolor icons.
51	4. Scaffold, don't write: `drawio-ai scaffold --list`, pick the closest template, then
52	`drawio-ai scaffold <name>.mjs -o <dir>/build.mjs` — the script arrives runnable and
53	self-checking. Edit only the deltas; Write a new script only if no template is close
54	AND you'd change more than half of it.
55	5. Each `node build.mjs` run prints validate JSON AND the render's machine-readable
56	`issues` list. Fix from THAT checklist — all issues in one Edit round — then re-run.
57	Loop until issues is empty.
58	6. Only when issues is empty: Read the PNG once as final visual confirmation. Target <= 2
59	PNG reads total. Then render once WITHOUT --check for the final deliverable PNG.
60	Do NOT invoke any drawio skill — this prompt already contains the full procedure.
61	Do not ask questions — make the standard choice and record it under ASSUMPTIONS.
62	Return EXACTLY this block, nothing else:
63	DRAWIO: <absolute path to .drawio>
64	PNG: <absolute path to .png>
65	VALIDATE: <verbatim final validate JSON>
66	ICONS: <comma-separated icon names used>
67	ITERATIONS: <number of render/fix cycles>
68	SUMMARY: <one sentence describing the diagram>
69	ASSUMPTIONS: <choices made without asking, or "none">
70	```
71
72	Relay `DRAWIO`, `PNG` and `SUMMARY` to the user verbatim; do NOT re-read the
73	.drawio or PNG in this conversation — the subagent already ran the vision
74	self-check. If `VALIDATE` is not ok, take over via the Inline path (the build
75	.mjs and .drawio are on disk at the returned paths).
76
77	## Inline path (no subagent support)
78
79	### 1. Shared Workflow
80
81	```bash
82	drawio-ai workflow
83	```
84
85	Prints the build → validate → render → write-to-project-path loop every diagram
86	follows. Read it; it is the source of truth for the process.
87
88	### 2. Domain rules
89
90	```bash
91	drawio-ai principles --mode databricks
92	```
93
94	Returns the Databricks rules + shared principles + catalog categories.
95
96	### 3. Build with the engine, then validat

People who liked this also viewed

6 picks

Type