agentleFS
probabl-aiprobabl-ai/skore-skills/skills/search-ml-literature/SKILL.md

search-ml-literature

Skill145 starsChanged 7 days agoBSD-3-ClauseView on GitHub

Search scientific and technical sources and write one idea file and one Ideas row for the direction the user confirms. Comes after review, beside the backlog. Do not call research-ml-practice. Do not write History, Backlog, or a design note. Do not pick a winner.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 4 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run python -m skore_skills status.

    line 27
  2. Distinctive:

    Search scientific and technical sources.

    line 35
  3. Very specific:

    Write scratch/research/direction-<slug>.md when that write can run.

    line 39
  4. Distinctive:

    If this turn is not already an explicit yes, restate the picked direction in one short paragraph, then one single-choice AskUserQuestion: …

    line 46
  5. Very specific:

    Do not edit History, Backlog, or Status.

    line 55
  6. Very specific:

    Do not load research-ml-practice.

    line 80
  7. Very specific:

    Do not write History, Backlog, Status, or a design note.

    line 81
  8. Very specific:

    Do not append a B<N> row.

    line 83
  9. Distinctive:

    Do not search a dataset proper name, a loader, or a Kaggle slug.

    line 84
  10. Very specific:

    Do not add a package or run env add / pixi add.

    line 85
  11. Distinctive:

    Do not treat one blog as ground truth.

    line 86

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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

  1. 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.
  2. 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.
  3. 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.
  4. Write scratch/research/direction-<slug>.md when that write can run. Do not paste it into chat.
  5. 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.
  6. AskUserQuestion: pick one, narrow the query, or stop. Narrow restarts at step 2. Stop writes no idea file.
  7. 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.
  8. 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.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.