Search ML Literature
Comes after review, beside the backlog. Source is a query, not an
artifact the user already named. Show a few directions, then write
one journal/ideas/<slug>.md and one Ideas row for the direction
they confirm. manage-ml-backlog triages that file later.
Human-facing prose
Details: setup-workspace references/human_facing_prose.md.
Directions and the pick question are scientific options for this
project — not skill ids or the wrapper CLI.
Procedure
- Run
python -m skore_skills status. Read journal/JOURNAL.md
Status, the last History headline, and the EDA summary when
data_analysis/data_analysis.md exists. If the query is too
vague to search, ask once.
- Rewrite the query as a problem class before any search. Drop
dataset names, loader names (
sklearn.datasets, fetch_*),
and Kaggle slugs. JOURNAL and EDA are context for that
rewrite, not search strings.
- Search scientific and technical sources. Fetch the primary
pages. When sources disagree, say so. Do not fill a gap from
memory after a search has run. If search cannot run, say so
and stop. Do not invent papers, URLs, or claims.
- Write
scratch/research/direction-<slug>.md when that write
can run. Do not paste it into chat.
- Show two to four directions. Each one: a one-line idea, why
it might transfer here, what it would change, and the URL
plus the claim. Do not rank a winner.
- AskUserQuestion: pick one, narrow the query, or stop.
Narrow restarts at step 2. Stop writes no idea file.
- If this turn is not already an explicit yes, restate the
picked direction in one short paragraph, then one
single-choice AskUserQuestion: Yes / No. Do not
also ask for a typed yes. Until Yes, that paragraph and the
question are the whole message. "Maybe" is not Yes. No or
Stop writes nothing.
- On Yes — the tool answer, or an explicit yes already given
on this or a later turn — write
journal/ideas/<slug>.md
and upsert its Ideas row. Do not create a design note.
Do not edit History, Backlog, or Status. Write
journal/ideas/<slug>.md:
# <slug>
- Experiment: <last History stem, or n/a>
- Source: literature: <url>
- Triage: open
- Question:
- Why now:
- What changes:
- Open gaps:
Open gaps hold transfer risks, disagreements, and domain
claims this dataset does not establish. A library named only
by a source stays in Open gaps until the user asks to add it.
Then upsert one ## Ideas row: Question as plain text, Status
open, Experiment <last History stem, or n/a>, Source
literature: <url>. If that table is missing, insert it
between History and Backlog. Do not rewrite the other
sections. Return the path to the caller.
Stop conditions
- Do not load
research-ml-practice.
- Do not write History, Backlog, Status, or a design note.
The Ideas table is the only
JOURNAL.md edit.
- Do not append a
B<N> row.
- Do not search a dataset proper name, a loader, or a Kaggle slug.
- Do not add a package or run
env add / pixi add.
- Do not treat one blog as ground truth.
1---2name: search-ml-literature3description: >4 Search scientific and technical sources and write one idea file and one5 Ideas row for the direction the user confirms. Comes after review, beside6 the backlog. Do not call research-ml-practice. Do not write History,7 Backlog, or a design note. Do not pick a winner.8metadata:9 source: https://skillmd.com/skills/probabl-ai/search-ml-literature10 modelTier: medium11---1213# Search ML Literature1415Comes after review, beside the backlog. Source is a query, not an16artifact the user already named. Show a few directions, then write17one `journal/ideas/<slug>.md` and one Ideas row for the direction18they confirm. `manage-ml-backlog` triages that file later.1920## Human-facing prose2122Details: `setup-workspace` `references/human_facing_prose.md`.23Directions and the pick question are scientific options for this24project — not skill ids or the wrapper CLI.2526## Procedure27281. Run `python -m skore_skills status`. Read `journal/JOURNAL.md`29 Status, the last History headline, and the EDA summary when30 `data_analysis/data_analysis.md` exists. If the query is too31 vague to search, ask once.322. Rewrite the query as a problem class before any search. Drop33 dataset names, loader names (`sklearn.datasets`, `fetch_*`),34 and Kaggle slugs. JOURNAL and EDA are context for that35 rewrite, not search strings.363. Search scientific and technical sources. Fetch the primary37 pages. When sources disagree, say so. Do not fill a gap from38 memory after a search has run. If search cannot run, say so39 and stop. Do not invent papers, URLs, or claims.404. Write `scratch/research/direction-<slug>.md` when that write41 can run. Do not paste it into chat.425. Show two to four directions. Each one: a one-line idea, why43 it might transfer here, what it would change, and the URL44 plus the claim. Do not rank a winner.456. **AskUserQuestion**: pick one, narrow the query, or stop.46 Narrow restarts at step 2. Stop writes no idea file.477. If this turn is not already an explicit yes, restate the48 picked direction in one short paragraph, then one49 single-choice **AskUserQuestion**: **Yes** / **No**. Do not50 also ask for a typed yes. Until Yes, that paragraph and the51 question are the whole message. "Maybe" is not Yes. No or52 Stop writes nothing.538. On Yes — the tool answer, or an explicit yes already given54 on this or a later turn — write `journal/ideas/<slug>.md`55 and upsert its Ideas row. Do not create a design note.56 Do not edit History, Backlog, or Status. Write57 `journal/ideas/<slug>.md`:5859```60# <slug>61- Experiment: <last History stem, or n/a>62- Source: literature: <url>63- Triage: open64- Question:65- Why now:66- What changes:67- Open gaps:68```6970 Open gaps hold transfer risks, disagreements, and domain71 claims this dataset does not establish. A library named only72 by a source stays in Open gaps until the user asks to add it.73 Then upsert one `## Ideas` row: Question as plain text, Status74 `open`, Experiment `<last History stem, or n/a>`, Source75 `literature: <url>`. If that table is missing, insert it76 between History and Backlog. Do not rewrite the other77 sections. Return the path to the caller.7879## Stop conditions8081- Do not load `research-ml-practice`.82- Do not write History, Backlog, Status, or a design note.83 The Ideas table is the only `JOURNAL.md` edit.84- Do not append a `B<N>` row.85- Do not search a dataset proper name, a loader, or a Kaggle slug.86- Do not add a package or run `env add` / `pixi add`.87- Do not treat one blog as ground truth.