n8n-goal-loop

Name: n8n-goal-loop
Author: buluslan

39stars2forks

How to install n8n-goal-loop?

buluslan/n8n-goal-loop/n8n-goal-loop

$npx -y skills add buluslan/n8n-goal-loop --skill n8n-goal-loop

Installs into the current project.

Summary

Generate copy-ready /goal commands that drive an agent through the full build→validate→deploy→layered-test→iterate loop for n8n workflows. Use when the user wants to build, modify, debug, or test an n8n workflow; needs a goal instruction or spec for n8n development; designs n8n n

Files · 1

View on GitHub

SKILL.md

1	# n8n Goal Loop
2
3	Turn an n8n workflow need into a copy-ready `/goal` that drives an agent through a full closed loop: build → validate → deploy → layered test → iterate.
4
5	This skill only generates the goal. It does not build, deploy, or run the workflow itself.
6
7	## Operating Mode
8
9	n8n workflows are complex. Align with the user first via a short interview, then generate — do not emit a goal from a single vague line (unless the user already states goal, node pipeline, and deliverable clearly).
10
11	Defaults:
12	- The user wants a paste-ready `/goal`, not a generic prompt.
13	- Keep the command prefix as `/goal` (not `/目标`).
14	- The first goal block must be the best recommended executable version, never a half-filled template — users copy the first block.
15	- For Chinese users: output the Chinese recommended version first, then an English-compatible mirror.
16	- Interview with numbered choices plus open questions. Never make the user fill out a long form.
17	- Ask only when the answer changes cost, risk, or direction. Auto-fill everything that has a standard n8n best practice (data-flow format, error handling, n8n red-lines) — do not ask about those.
18	- Every n8n goal must separate script-level testing from end-to-end testing. `execution success` is not proof the workflow works for the business.
19	- Generic categories only — never hardcode a specific vendor, brand, or the user's private project details.
20
21	## Workflow
22
23	1. Classify the task: build new workflow / modify existing / debug a failure / test & verify.
24	2. Run the alignment interview (see `references/interview-checklist.md`). Four layers:
25	- Goal & deliverable — what problem, what final artifact.
26	- Node pipeline — data source → processing → output sink.
27	- Selection (generic categories) — trigger / deployment / data source / output channel.
28	- Red lines — what must not be touched.
29	3. Generate the nine-element goal (see `references/n8n-9-elements.md` and `references/goal-command-playbook.md`).
30	4. Let the user confirm or adjust by replying with numbered choices (e.g. `按默认` or `1A 2C`).
31	5. Check the goal against `references/n8n-pitfalls.md` so known traps are pre-empted.
32	6. For file deliverables, run `python3 scripts/lint_goal_command.py <file>` before calling the goal done.
33
34	## Output Contract
35
36	Nine elements (qiaomu seven + data-flow + error handling). See `references/n8n-9-elements.md` for full detail.
37
38	```text
39	推荐执行版（中文，可直接复制）
40
41	/goal [Outcome: node pipeline 输入→处理→输出 + deliverable].
42	数据流：[node-to-node field mapping + key input/output fields + data format].
43	错误处理：[failure strategy for key nodes: continueOnFail / fallback store / degrade].
44	验证：[static check → deploy visible → script test (structure/start/execution progress) → end-to-end test (real data, execution success does not count)].
45	约束：[n8n red lines: no DB damage / no hardcoded secrets / Code node no HTTP / jsonBody UTF-8 / AI must not fabricate data].
46	边界：[only the target workflow, no others].
47	迭代策略：[small sample first; refresh cache after PUT; fix against pitfall library; max 3 rounds].
48	完成条件：[end-to-end test passes + business fields have real values].
49	暂停条件：[real credentials / publish-activate / real money / instance anomaly (e.g. SQLITE_CORRUPT)].
50
51	默认选择理由：[one sentence].
52
53	可选调整
54	1. [dimension]: A (default) / B / C
55
56	你可以直接回复：按默认，或类似 1A 2C.
57
58	Goal Draft (English-compatible)
59	[faithful English mirror with the same nine elements]
60	```
61
62	## Quality Bar
63
64	A strong n8n goal:
65	- Outcome states the node pipeline (input → process → output) and the deliverable.
66	- Data-flow contract covers node-to-node field mapping and key fields.
67	- Error handling covers failure fallback for fragile nodes.
68	- Verification separates script test from end-to-end test, with real business evidence.
69	- Constraints include n8n red lines (no DB damage, no hardcoded secrets, Code node does not issue HTTP, jsonBody uses UTF-8).
70	- Pause conditions cover real credentials, publish/activate, real money, and instance anomalies.
71
72	Reject or revise a goal that:
73	- Outcome says only "build a workflow" with no pipeline.
74	- Verification says only "test passes" with no layering.
75	- Missing data-flow contract or error handling.
76	- Uses `execution success` as completion proof.
77	- Hardcodes a specif

People who liked this also viewed

6 picks

Type

n8n-goal-loop

39stars2forks

View on GitHub

How to install n8n-goal-loop?

buluslan/n8n-goal-loop/n8n-goal-loop

$npx -y skills add buluslan/n8n-goal-loop --skill n8n-goal-loop

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# n8n Goal Loop
2
3	Turn an n8n workflow need into a copy-ready `/goal` that drives an agent through a full closed loop: build → validate → deploy → layered test → iterate.
4
5	This skill only generates the goal. It does not build, deploy, or run the workflow itself.
6
7	## Operating Mode
8
9	n8n workflows are complex. Align with the user first via a short interview, then generate — do not emit a goal from a single vague line (unless the user already states goal, node pipeline, and deliverable clearly).
10
11	Defaults:
12	- The user wants a paste-ready `/goal`, not a generic prompt.
13	- Keep the command prefix as `/goal` (not `/目标`).
14	- The first goal block must be the best recommended executable version, never a half-filled template — users copy the first block.
15	- For Chinese users: output the Chinese recommended version first, then an English-compatible mirror.
16	- Interview with numbered choices plus open questions. Never make the user fill out a long form.
17	- Ask only when the answer changes cost, risk, or direction. Auto-fill everything that has a standard n8n best practice (data-flow format, error handling, n8n red-lines) — do not ask about those.
18	- Every n8n goal must separate script-level testing from end-to-end testing. `execution success` is not proof the workflow works for the business.
19	- Generic categories only — never hardcode a specific vendor, brand, or the user's private project details.
20
21	## Workflow
22
23	1. Classify the task: build new workflow / modify existing / debug a failure / test & verify.
24	2. Run the alignment interview (see `references/interview-checklist.md`). Four layers:
25	- Goal & deliverable — what problem, what final artifact.
26	- Node pipeline — data source → processing → output sink.
27	- Selection (generic categories) — trigger / deployment / data source / output channel.
28	- Red lines — what must not be touched.
29	3. Generate the nine-element goal (see `references/n8n-9-elements.md` and `references/goal-command-playbook.md`).
30	4. Let the user confirm or adjust by replying with numbered choices (e.g. `按默认` or `1A 2C`).
31	5. Check the goal against `references/n8n-pitfalls.md` so known traps are pre-empted.
32	6. For file deliverables, run `python3 scripts/lint_goal_command.py <file>` before calling the goal done.
33
34	## Output Contract
35
36	Nine elements (qiaomu seven + data-flow + error handling). See `references/n8n-9-elements.md` for full detail.
37
38	```text
39	推荐执行版（中文，可直接复制）
40
41	/goal [Outcome: node pipeline 输入→处理→输出 + deliverable].
42	数据流：[node-to-node field mapping + key input/output fields + data format].
43	错误处理：[failure strategy for key nodes: continueOnFail / fallback store / degrade].
44	验证：[static check → deploy visible → script test (structure/start/execution progress) → end-to-end test (real data, execution success does not count)].
45	约束：[n8n red lines: no DB damage / no hardcoded secrets / Code node no HTTP / jsonBody UTF-8 / AI must not fabricate data].
46	边界：[only the target workflow, no others].
47	迭代策略：[small sample first; refresh cache after PUT; fix against pitfall library; max 3 rounds].
48	完成条件：[end-to-end test passes + business fields have real values].
49	暂停条件：[real credentials / publish-activate / real money / instance anomaly (e.g. SQLITE_CORRUPT)].
50
51	默认选择理由：[one sentence].
52
53	可选调整
54	1. [dimension]: A (default) / B / C
55
56	你可以直接回复：按默认，或类似 1A 2C.
57
58	Goal Draft (English-compatible)
59	[faithful English mirror with the same nine elements]
60	```
61
62	## Quality Bar
63
64	A strong n8n goal:
65	- Outcome states the node pipeline (input → process → output) and the deliverable.
66	- Data-flow contract covers node-to-node field mapping and key fields.
67	- Error handling covers failure fallback for fragile nodes.
68	- Verification separates script test from end-to-end test, with real business evidence.
69	- Constraints include n8n red lines (no DB damage, no hardcoded secrets, Code node does not issue HTTP, jsonBody uses UTF-8).
70	- Pause conditions cover real credentials, publish/activate, real money, and instance anomalies.
71
72	Reject or revise a goal that:
73	- Outcome says only "build a workflow" with no pipeline.
74	- Verification says only "test passes" with no layering.
75	- Missing data-flow contract or error handling.
76	- Uses `execution success` as completion proof.
77	- Hardcodes a specif

People who liked this also viewed

6 picks

Type