test-ml-pipeline

Name: test-ml-pipeline
Author: probabl-ai

70stars4forks

How to install test-ml-pipeline?

probabl-ai/skills/test-ml-pipeline

$npx -y skills add probabl-ai/skills --skill test-ml-pipeline

Installs into the current project.

Summary

Owns the tests/ folder of an ML workspace and the pairing rule between an experiment and its tests. Lightweight router: every test category has its own subskill (smoke-test-ml-pipeline is the only one for v1; regression-test-ml-pipeline / distribution-test-ml-pipeline etc

Files · 1

View on GitHub

SKILL.md

1	# Test ML Pipeline (router)
2
3	Where tests for an ML workspace live, what gets paired with what,
4	and which subskill owns the body of each test category.
5
6	## First action (every turn)
7
8	Before answering anything else:
9
10	1. Confirm an approved design note exists for the test
11	the user is asking about. The pairing rule is hard:
12	`tests/<category>/test_NN_<short_name>.py` only exists if
13	`journal/NN_<short_name>.md` is at least `approved` and
14	`experiments/NN_<short_name>.py` is the matching script. If
15	the design note doesn't exist, hand back to `iterate-ml-experiment`.
16	2. Emit the Pre-flight checklist (below) as visible text in
17	your response, with each box marked.
18	3. Use the Dispatch table to pick the subskill that owns the
19	test category, then hand off.
20
21	## Pre-flight — emit this checklist as visible text before any test work
22
23	```
24	Pre-flight (test-ml-pipeline):
25	- [ ] `journal/NN_<short_name>.md` exists and is at least `approved`
26	(or confirmed n/a — about to hand off to `iterate-ml-experiment`)
27	- [ ] `experiments/NN_<short_name>.py` exists with the matching stem
28	(or confirmed n/a — about to hand off to `organize-ml-workspace`)
29	- [ ] Test category picked: smoke \| regression \| distribution \| …
30	- [ ] Subskill dispatched: `smoke-test-ml-pipeline` \| …
31	- [ ] Test file stem decided: `tests/<category>/test_NN_<short_name>.py`
32	- [ ] pytest is on the project's dependency manifest (per
33	`data-science-python-stack` § Tier 1)
34	```
35
36	## Stop conditions — read before anything else
37
38	- No test without an approved design note. Never create
39	`tests/<category>/test_NN_.py` if the matching `journal/NN_.md`
40	isn't on disk and at least `approved`. The design note is the contract;
41	the test asserts the contract holds. Reverse order is incoherent.
42	- The stem rule is hard. Test file basename is
43	`test_NN_<short_name>.py` (with the `test_` prefix that pytest
44	expects); the `NN_<short_name>` portion matches the experiment
45	exactly. One experiment → one test file per category. No
46	`test_<NN>_v2.py`, no `test_NN_<short_name>_2.py`. If a test
47	needs to evolve, edit it in place; the pairing must stay 1:1.
48	- One subskill per category. This skill only places the empty
49	test file and hands off. Don't write assertion bodies, fixture
50	construction, or test-specific logic in this skill — that belongs
51	to the matching subskill (`smoke-test-ml-pipeline`, etc.).
52	- pytest is the runner. Tests are pytest tests, not
53	jupytext-style `# %%` scripts. The experiment scripts live in
54	`experiments/` and stay `# %%`-style for interactive iteration;
55	the tests are binary pass/fail and benefit from pytest's
56	reporting. Don't mix the two conventions.
57
58	## Layout this skill owns
59
60	```
61	project/
62	└── tests/
63	├── smoke/
64	│ ├── test_01_baseline.py # ↔ experiments/01_baseline.py
65	│ ├── test_02_short_name.py
66	│ └── ...
67	└── (future: regression/, distribution/, …)
68	```
69
70	The pairing rule:
71
72	```
73	journal/NN_<short_name>.md
74	experiments/NN_<short_name>.py
75	tests/<category>/test_NN_<short_name>.py
76	```
77
78	— same `NN_<short_name>` stem in all three. The `test_` prefix on
79	the test file basename is the pytest naming

People who liked this also viewed

6 picks

Type

test-ml-pipeline

70stars4forks

View on GitHub

How to install test-ml-pipeline?

probabl-ai/skills/test-ml-pipeline

$npx -y skills add probabl-ai/skills --skill test-ml-pipeline

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Test ML Pipeline (router)
2
3	Where tests for an ML workspace live, what gets paired with what,
4	and which subskill owns the body of each test category.
5
6	## First action (every turn)
7
8	Before answering anything else:
9
10	1. Confirm an approved design note exists for the test
11	the user is asking about. The pairing rule is hard:
12	`tests/<category>/test_NN_<short_name>.py` only exists if
13	`journal/NN_<short_name>.md` is at least `approved` and
14	`experiments/NN_<short_name>.py` is the matching script. If
15	the design note doesn't exist, hand back to `iterate-ml-experiment`.
16	2. Emit the Pre-flight checklist (below) as visible text in
17	your response, with each box marked.
18	3. Use the Dispatch table to pick the subskill that owns the
19	test category, then hand off.
20
21	## Pre-flight — emit this checklist as visible text before any test work
22
23	```
24	Pre-flight (test-ml-pipeline):
25	- [ ] `journal/NN_<short_name>.md` exists and is at least `approved`
26	(or confirmed n/a — about to hand off to `iterate-ml-experiment`)
27	- [ ] `experiments/NN_<short_name>.py` exists with the matching stem
28	(or confirmed n/a — about to hand off to `organize-ml-workspace`)
29	- [ ] Test category picked: smoke \| regression \| distribution \| …
30	- [ ] Subskill dispatched: `smoke-test-ml-pipeline` \| …
31	- [ ] Test file stem decided: `tests/<category>/test_NN_<short_name>.py`
32	- [ ] pytest is on the project's dependency manifest (per
33	`data-science-python-stack` § Tier 1)
34	```
35
36	## Stop conditions — read before anything else
37
38	- No test without an approved design note. Never create
39	`tests/<category>/test_NN_.py` if the matching `journal/NN_.md`
40	isn't on disk and at least `approved`. The design note is the contract;
41	the test asserts the contract holds. Reverse order is incoherent.
42	- The stem rule is hard. Test file basename is
43	`test_NN_<short_name>.py` (with the `test_` prefix that pytest
44	expects); the `NN_<short_name>` portion matches the experiment
45	exactly. One experiment → one test file per category. No
46	`test_<NN>_v2.py`, no `test_NN_<short_name>_2.py`. If a test
47	needs to evolve, edit it in place; the pairing must stay 1:1.
48	- One subskill per category. This skill only places the empty
49	test file and hands off. Don't write assertion bodies, fixture
50	construction, or test-specific logic in this skill — that belongs
51	to the matching subskill (`smoke-test-ml-pipeline`, etc.).
52	- pytest is the runner. Tests are pytest tests, not
53	jupytext-style `# %%` scripts. The experiment scripts live in
54	`experiments/` and stay `# %%`-style for interactive iteration;
55	the tests are binary pass/fail and benefit from pytest's
56	reporting. Don't mix the two conventions.
57
58	## Layout this skill owns
59
60	```
61	project/
62	└── tests/
63	├── smoke/
64	│ ├── test_01_baseline.py # ↔ experiments/01_baseline.py
65	│ ├── test_02_short_name.py
66	│ └── ...
67	└── (future: regression/, distribution/, …)
68	```
69
70	The pairing rule:
71
72	```
73	journal/NN_<short_name>.md
74	experiments/NN_<short_name>.py
75	tests/<category>/test_NN_<short_name>.py
76	```
77
78	— same `NN_<short_name>` stem in all three. The `test_` prefix on
79	the test file basename is the pytest naming

People who liked this also viewed

6 picks

Type