agentleFS
lenar-amirovlenar-amirov/product-pipeline-public/.claude/skills/ingest/SKILL.md

ingest

Skill13 starsChanged 5 months agoMITView on GitHub

First-class intake of external data the PM was given — a deck (PDF/PPTX), analytics export (xlsx/csv), Confluence page, or pasted numbers. Extracts key metrics with exact locations, maps them onto open hypotheses in the registry, proposes upgrades/refutations. Use when the PM says "мне принесли дек", "вот данные/выгрузка", "прочитай этот PDF", "ingest this", "here's a deck from the analyst".

90 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use daily-arxiv, exp-eval and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • phanghonghaoSkill

    poster

    HTML/CSS A4 海报生成。从沙龙/活动 Markdown 文件生成科技风 A4 海报 PDF/PNG。Use when user mentions: poster, 海报, 海报生成, make poster, generate poster, 活动海报, 沙龙海报.

    Also in 5 projects that use this

  • skyllwtSkill

    exp-pilot-eval

    Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.

    Also in 4 projects that use this

  • skyllwtSkill

    exp-pilot-run

    Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.

    Also in 4 projects that use this

  • +6 more from Lambenthan
  • +1 more from skyllwt

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 74 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    xlsx/csv: read directly (xlsx skill if available).

    line 20
  2. Distinctive:

    If no initiative exists: do steps 1–2 and 5 in-chat, then offer to persist (FIRST LAUNCH flow).

    line 72

“Common” is measured: kept by at least 6 in 10 of the 74 edited versions of this file we could read. “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

Ingest — /ingest <file | Confluence pageID | pasted data>

In real corporate work the most valuable data arrives through the side door — someone else's deck, an ad-hoc export, a wiki page. This job turns such an artifact into typed evidence in the registry instead of a manual retyping exercise.

1. Acquire

  • PDF (deck, report): render pages to images — python3 tools/scripts/render-pdf.py <file.pdf> <scratch_dir> [first] [last] — then Read the images (batch 5-10 pages per pass).
  • PPTX: convert to PDF first if possible (soffice --convert-to pdf), or ask the PM for a PDF export.
  • xlsx/csv: read directly (xlsx skill if available).
  • Confluence page: fetch via MCP (confluence_get_page with pageID).
  • Pasted text/numbers: work with it as-is.

2. Extract

Pull out every number and finding relevant to the initiative's tracks: metric, value, definition (exactly as the source states it), and location ("слайд 17", "стр. 3, таблица 2"). Note the data window (dates) — it goes into the source record.

3. Map onto hypotheses (the core of the job)

Load open hypotheses: python3 tools/scripts/hypotheses.py show <dir>. For each extracted finding, classify:

  • Supports Hx → propose upgrade: hypotheses.py set Hx --status confirmed --type REAL --confidence 0.N --add-source "<file>::слайд NN — <what it says>"
  • Contradicts Hx → propose refutation (--status refuted) and say explicitly that REAL beats the previous INFERRED/SYNTHETIC — document the delta in the narrative file.
  • Contradicts ANOTHER source already in the registry → do NOT silently average: hypotheses.py set Hx --flag data_inconsistency --note "source A vs source B: X vs Y" and add the reconciliation question to the open questions for the analyst.
  • Relevant but matches no hypothesis → candidate NEW hypothesis (hypotheses.py add) — ask the PM before adding.
  • Not mapped → append to research/inbox-notes.md with source refs, so nothing is lost.

4. Typing discipline

Internal analytics/dashboards/decks of your own product = REAL. External benchmarks and competitor numbers = INFERRED (0.3–0.5) — they are analogies, not facts about your product. Surveys of your users = REAL. Someone's opinion slide without data = AMBIGUOUS (0.1–0.3).

Source quality check before typing anything REAL: who produced it, what data window, does its total reconcile with numbers already in the registry? A REAL source with an unknown window or an unreconciled total enters at the bottom of the REAL range (0.6), not the top.

5. Deliver

Summary to the PM: what was ingested (N pages/rows), which hypotheses moved (before → after), which contradictions were flagged, what landed in inbox-notes, and the updated registry table (hypotheses.py render). Then: research/analytics-data.md (or a new file under research/) gets the extracted numbers with locations — future jobs cite it, not the raw artifact.

If no initiative exists: do steps 1–2 and 5 in-chat, then offer to persist (FIRST LAUNCH flow).

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.