gigaplexity-file-intelligence

name	gigaplexity-file-intelligence
description	File intelligence skill for understanding documents, images, and audio through Gigaplexity `ask` attachments. MUST use when user asks to analyze/compare/summarize local files (PDF, DOC, image, audio, code/text) with evidence-aware outputs.
argument-hint	Goal + absolute file paths + expected output format

Gigaplexity File Intelligence Skill

Purpose

Use this skill to extract reliable insights from user files (docs/images/audio), then produce concise, structured outputs aligned with the task (summary, Q&A, comparison, due diligence, extraction).

MUST Use When

Use this skill immediately if user intent includes any of these:

• “прочитай/разбери файл”, “что в этом PDF/DOC?”, “проанализируй картинку/аудио”.
• Need to summarize or compare attached/local files.
• Need factual extraction from file content (entities, dates, obligations, risks, action items).
• Need multimodal understanding (document + image, or image-only, audio-only).

Core Capability Constraint (critical)

Gigaplexity ask supports attachments, but all files in one request must be from the same category:

• Document/text/code files
• Images
• Audio

If user gives mixed categories, split into multiple passes and then synthesize.

Workflow

1) Input Validation

• Ensure file paths are absolute and files exist.
• Determine category for each file.
• If categories mixed: batch by category and process separately.

2) Intent Clarification (if needed)

Ask minimal clarifying questions only when output target is ambiguous:

• summary vs extraction vs comparison vs risk review
• desired language and level (brief / detailed)

3) File Analysis via Gigaplexity

• Use #tool:gigaplexity/ask with file_paths.
• Provide a precise task prompt (what to extract, what to ignore, desired format).
• For large/complex files, run iterative passes:
  • pass A: global summary
  • pass B: targeted extraction
  • pass C: contradiction/risk scan

4) Verification Pass

• For high-stakes tasks (legal/finance/security), request explicit evidence snippets from file content.
• If content is unclear, state uncertainty and what additional file/context is needed.

5) Synthesis

• Merge per-file or per-batch findings.
• Highlight agreements, conflicts, missing info, and next actions.

Prompt Patterns (ready to use)

A) Document Summary

"Analyze attached document(s). Return:

1. 5-bullet executive summary,
2. key entities/dates/amounts,
3. obligations and risks,
4. open questions. Do not invent missing facts."

B) Image Analysis

"Describe attached image(s) in detail, then extract actionable facts relevant to: . Distinguish observable facts from assumptions."

C) Audio Analysis

"Transcribe and summarize attached audio. Return speakers (if inferable), key points, decisions, action items, timestamps when possible, and uncertainty notes."

D) Multi-file Comparison (same category)

"Compare attached files and return: common points, differences, contradictions, and recommendation."

Output Contract (default)

1. What was analyzed (files/categories)
2. Executive Summary
3. Extracted Facts (structured)
4. Risks / Contradictions / Gaps
5. Next Actions

Quality Rules

• Never claim a file says something unless model output supports it.
• Separate observed content from inferred interpretation.
• Keep outputs decision-friendly: short, structured, and actionable.
• If source quality is insufficient, explicitly say what is missing.

Anti-Patterns (strictly avoid)

• Mixing file categories in one ask call.
• Vague “looks good” summaries without concrete extracted facts.
• Ignoring ambiguities in OCR/audio or low-quality images.
• Overconfident conclusions in high-risk contexts.

Completion Checklist

• File category rule respected (single category per request)
• Output matches requested task type
• Facts and uncertainties clearly separated
• Contradictions/gaps explicitly listed
• Recommendation or next steps provided