zy839971925-zyy/Agent--deep-research-Workflow-Skills / skills/reasoning-workflow
reasoning-workflow
VerifiedDefault silent workflow for everyday questions, research, writing, design and engineering. Match effort to the request, resolve costly ambiguity, find specialist resources and verify the requested outcome. Explicitly selectable with “调用我的 workflow skills”.
references/multimodal-and-data.md
View on GitHub# Multimodal and data evidence
**Trigger:** Claims depend on images, charts, tables, datasets, audio, video, OCR/transcription, or calculations.
**Reads:** original modality/data, metadata, task claim, transformation/calculation requirements.
**Updates:** observations, derived records, calculations, transformation lineage, limitations.
**May invalidate:** claims based on transcription/OCR/chart-reading/calculation errors or transformed material mistaken for original.
**Exit:** relevant modality/data has been inspected/processed with lineage and uncertainty preserved.
**Related:** [evidence and provenance](evidence-and-provenance.md), [traceability and integrity](traceability-and-integrity.md), [retrieval and observation](retrieval-and-observation.md).
**Return:** [root workflow](../SKILL.md) → Evidence / Verification.
Use when conclusions depend on images, PDFs, charts, tables, spreadsheets, datasets, audio, video, OCR, metadata, or calculations.
## Observe the right modality
Extraction can lose layout, legends, footnotes, merged cells, visual emphasis, signs, units, or speaker/timing information. Inspect the original modality when those affect a material claim. If the modality cannot be inspected, narrow the claim instead of inventing details.
Retain useful locators: page/figure/table, slide, sheet/cell, image region, timestamp, file version, or code location. Distinguish original pixels/audio, OCR, transcripts, metadata, derived data, and secondary descriptions.
## Tables and charts
Check units, axes, log scales, denominators, frequency, sample scope, missing values, footnotes, revisions, and comparison periods. OCR may swap signs, decimals, labels, or columns. Prefer underlying data for exact quantities; label visual estimates.
A chart association does not establish causation. A generated illustration is not measurement evidence.
## Calculations and data transformations
Separate source inputs, transformations, assumptions, and derived results. For consequential numbers, retain reproducible formulas or computations and input locators. Align currency, date, nominal/real, stock/flow, unit, population, and category definitions.
Inspect missingness, selection, revisions, measurement changes, and overlap. Do not pool incompatible metrics or studies merely to obtain one number.
## Audio and video
A transcript can support spoken words within transcription limits, not visual actions. Sparse frames do not establish continuous behavior. Preserve consequential uncertainty about speaker, timing, cuts, or editing. Use native media capabilities where available instead of pretending a text transcript captures the full evidence.
## Data-generating process and selection path
Before treating a dataset as a window onto reality, ask how reality became the observed data:
`real event/population → measurement → collection → eligibility/selection → recording → cleaning/transformation → stored dataset → analysis`
A numerically correct analysis can still be substantively wrong when the selection path excludes the cases that matter. Inspect non-observation, non-reporting, survivorship, changed definitions, instrumentation changes, missing groups, duplicated units, and incentives that affect what gets recorded. Use this only when selection could change a material claim; do not turn every small table into a full causal data audit.
[← Return to root workflow](../SKILL.md)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |