Execute qualquer Skill no Manus
com um clique

Execute qualquer Skill no Manus com um clique

undermind-search

Estrelas1

Forks0

Atualizado22 de julho de 2026 às 09:34

Stage 1 of the lit review pipeline: run an Undermind.ai "Classic" deep search from the natural-language brief produced by Stage 0, then parse + enrich the exported references into the pipeline schema. The driver logs in automatically with stored credentials and runs headless via Playwright. Only use this skill when explicitly requested. Do NOT auto-trigger on general literature review or paper search requests.

Instalação

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Executar no Manus

Fonte

kennethkhoocy

kennethkhoocy/applied-micro-skills

Abrir repositório GitHub Ver repositórios do creator

Download

Executar no Manus

Ocupações relacionadasSOC

Baseado na classificação ocupacional SOC

Professores do ensino superior, todos os outrosEducação e Biblioteconomia·SOC 25-1199

Explorador de arquivos

3 arquivos

SKILL.md

readonly

Mais deste repositório

mesmo repositório

adjudication-sheets

kennethkhoocy/applied-micro-skills

Build human adjudication / hand-labeling sheets from LLM-pipeline data without evidence truncation. Use when: (1) preparing a CSV/Excel sheet for a human to rule on cases an LLM classifier or rater panel judged, (2) a labeler reports "there is no information to label from" or cells look empty in Excel, (3) excerpt columns cluster at one exact length (e.g. all 1,500 chars — a hard truncation cap). Covers: full rating-basis recovery, Excel 32,767-char cell cap, multi-line CSV mangling, ruling dropdowns, companion text files.

2026-07-221

adversarial-empirical-review

kennethkhoocy/applied-micro-skills

N-round adversarial review pipeline for empirical research output — the chain from data to LaTeX tables to a manuscript that cites them. A Claude drafter proposes minimal diffs, a deterministic mechanical battery gates every diff from a clean state with a regression gate, a Codex reviewer files check-backed critiques, and a blind judge panel decides residual disputes. Manual-invoke ONLY: trigger when the user explicitly runs /adversarial-empirical-review or names 'adversarial-empirical-review' / 'adversarial empirical review'. Do NOT auto-trigger on generic 'review my results', 'check my tables', or manuscript-editing requests. For prose-style refinement use style-emulation instead; this skill AUDITS WHETHER THE TABLES ARE CORRECT — that each number in the tables is what the analysis code computes, reproduces from the data, and is internally consistent. It is an empirical + code review: the manuscript is read only to resolve table numbering, and prose is not examined.

2026-07-221

annotator-input-parity-check

kennethkhoocy/applied-micro-skills

Before designing, training, or auditing ANY model that replicates human-annotated labels, audit the annotation protocol's INPUT — the exact document/evidence the human labelers consulted — and give the model that same input. Use when: (1) designing a classifier/LLM extractor whose target is a hand-coded label set, (2) a label-replication model shows low recall concentrated in a label subset and the diagnosis on offer is "the label's information is not in the features", (3) reviewers propose construct splits (e.g. "designation vs record-evident"), adjudication sittings, or per-domain stop rules to explain residual disagreement with gold, (4) validating an extraction pipeline against labels transcribed from a source document. Symptom of the underlying failure: elaborate theory accumulates to explain why gold is "partially unpredictable" when the model was simply never shown the document the annotators read.

2026-07-221

asyncopenai-concurrency-httpx-pool

kennethkhoocy/applied-micro-skills

Raise real concurrency in asyncio LLM batch scorers built on the OpenAI SDK (AsyncOpenAI, including OpenAI-compatible providers like DeepSeek). Use when: (1) raising an asyncio.Semaphore above ~100 produces no throughput gain, (2) a batch pipeline saturates near 100 in-flight requests despite a larger semaphore, (3) planning a high-concurrency campaign against a provider with no hard rate limit (DeepSeek v4-flash tolerates 2000+ in flight). Root cause: AsyncOpenAI's default httpx pool caps max_connections at 100, silently bottlenecking any larger semaphore — you must pass a custom http_client with httpx.Limits sized to the semaphore.

2026-07-221

cite-placement

kennethkhoocy/applied-micro-skills

Place pre-screened literature citations into a LaTeX or Word manuscript, or restyle the citations already in one. Three modes: (1) inline placement — inline \cite{}/\citet{}/\citep{} with a compiled references.bib, for author-date journals (APA, MLA, Harvard, Chicago author-date, IEEE, Vancouver); (2) footnote placement — full formatted \footnote{} or OOXML footnotes for legal and notes styles (Bluebook, OSCOLA, Chicago, APA, McGill) with Id./supra short forms; (3) restyle — convert existing footnote citations from one style to another. This skill is manual-invoke ONLY — trigger ONLY when the user explicitly runs /cite-placement or explicitly names the "cite-placement" skill. Do NOT auto-trigger on general citation, footnote, or reference requests.

2026-07-221

download-gated-pdfs

kennethkhoocy/applied-micro-skills

Download the actual PDF binary from bot-gated sites (taxpolicycenter.org, urban.org, SSRN-hosted mirrors, think-tank/publisher sites) via the Wayback Machine id_ URL form. Use when: (1) curl/WebFetch of a .pdf URL returns HTML instead of a PDF even with a browser User-Agent, (2) pypdf fails with "invalid pdf header: b'<!DOC'" or "EOF marker not found" on a freshly downloaded file, (3) Firecrawl can parse the PDF to markdown but you need the original file on disk (e.g., filing a reference copy).

2026-07-221

name

undermind-search

description

Undermind Search (Stage 1)

Takes the Undermind brief that Stage 0 extracts and returns the papers Undermind finds, enriched and saved as <stem>.json + <stem>.bib for dedup and screening.

The stage is two halves:

scripts/undermind_search.py — a Playwright driver that signs in, drives the Classic search UI, and exports the references (BibTeX by default).
scripts/undermind_ingest.py — UI-independent parsing + enrichment. It reads the exported .bib/.ris, fills missing DOIs (Crossref) and abstracts/journals (OpenAlex), and writes the pipeline JSON. Importable, and runnable standalone on any reference file.

Quick start

# One-time setup: store Undermind credentials and verify sign-in (opens a browser)
python scripts/undermind_search.py --login

# Driven by the orchestrator (the normal path)
python scripts/undermind_search.py --brief-file undermind_brief.txt \
    -o stage1_undermind.json --debug-dir debug_undermind

# Standalone, from a short query, watching the browser
python scripts/undermind_search.py --query "dual-class shares cost of equity" --headed

# Ingest an already-exported file (no browser)
python scripts/undermind_ingest.py --input references.bib -o stage1_undermind.json

The flow (validated against the June 2026 UI)

app.undermind.ai → (auto-login if needed) → sidebar Classic → Search → type the brief into "I want to find…" → answer Undermind's clarifying questions (a Sonnet subagent by default, Opus when all_opus, via --answers-dir; or the Sonnet API as the standalone fallback) → Generate Research Report → wait for the report → Export → BibTeX → parse + enrich → stage1_undermind.json (+ .bib).

Undermind's clarifying questions render as clickable chips, but the composer also accepts free text; the driver answers in free text, which is robust to UI changes.

Login & credentials

The driver uses a dedicated persistent browser profile (--user-data-dir, default ~/.undermind-profile) and signs in with:

Variable	Meaning
`UNDERMIND_EMAIL`	Undermind account email
`UNDERMIND_PASSWORD`	Undermind account password

Stored in ~/.lit-review-pipeline.env (gitignored; never commit credentials). --login captures them on first run if they are absent (prompts, then writes them to the env file) and verifies sign-in. Once the profile holds a valid session, later runs skip the login step; if the session lapses, the driver logs in again from the stored credentials.

Terms-of-service note. This stage automates a logged-in browser session against Undermind, a paid deep-search service with no public API, using your own paid account. Automated access of this kind may fall outside Undermind's terms of service, so you run the stage at your own risk and under your own account. The orchestrator runs without it — --skip undermind (or --quick) skips it and the remaining channels carry the run — so the stage is optional.

Answering clarifying questions — two modes. In the orchestrator's agent-driven flow, pass --answers-dir DIR: the driver writes each question to clarify_request_<n>.json and waits for the orchestrator to drop clarify_answer_<n>.json ({"answer": "..."}) — produced by a Sonnet subagent by default, or Opus when all_opus — then types it — no API key is used, and the stage is interactive in this mode. Without --answers-dir (standalone or autonomous fallback) the driver answers with the Anthropic API (Sonnet), which requires ANTHROPIC_API_KEY.

CLI

Flag	Default	Description
`--brief-file PATH`	—	File holding the brief (the orchestrator passes this)
`--query TEXT`	—	Short brief for standalone use
`-o, --output PATH`	`stage1_undermind.json`	Output JSON (a `.bib` sibling is written)
`--debug-dir PATH`	—	Save step screenshots / HTML dumps here
`--format {bibtex,ris}`	`bibtex`	Format to export from Undermind
`--user-data-dir PATH`	`~/.undermind-profile`	Persistent browser profile
`--model ID`	`claude-sonnet-4-6`	Model for clarifying answers in API/autonomous mode
`--answers-dir PATH`	—	Agent-in-the-loop: hand clarifying questions to a subagent via files (no API), Sonnet by default / Opus when `all_opus`; the orchestrator default
`--headed`	off	Show the browser (debugging)
`--login`	off	First-run setup: capture credentials + verify sign-in (headed)
`--no-enrich`	off	Skip Crossref/OpenAlex enrichment

Graceful degradation

If credentials are missing, login fails, or the export does not produce a file, the driver prints the UNDERMIND_DEFERRED sentinel, writes an empty result list, and exits 0. The orchestrator marks the stage deferred and the rest of the pipeline (Google Scholar, dedup, screening) still completes. The brief is always available in undermind_brief.txt for a manual run.

Output schema

Same as the other stages, with source: "undermind":

{"title": "...", "authors": "A, B", "year": "2024",
 "doi": "https://doi.org/10.x/y", "abstract": "...", "journal": "...",
 "url": "...", "source": "undermind", "verified": true,
 "citations": 0, "open_access": false}

Testing with windows-mcp (headed)

The Playwright driver is the production path. When the Undermind UI changes and a selector needs re-confirming, drive the live site by hand with the windows-mcp desktop tools (open Chrome, walk Classic → Search → export) to read the current element labels/roles, then update the corresponding Playwright locators here. The durable locators in use — roles/text/placeholders ("Classic", "Search", placeholder "I want to find…", "Generate Research Report", "Export", menu item "BibTeX", inputs #email/#password) — are chosen to survive minor UI churn.

Troubleshooting

UNDERMIND_DEFERRED: Not logged in... — run --login once, or check the env credentials. Inspect --debug-dir screenshots (00_landing, 00a_login_modal).
Stuck after launch — wait_for_report polls up to 20 min. Check 05_search_launched / 06_report_* screenshots; the "Generate Research Report" or "Export" locator may need re-confirming via windows-mcp.
Login special characters — credentials are read from the env file (not the shell), so passwords with !, $, etc. are passed through unmangled.