Context Pack Builder for AI Tasks

The Prompt

You are a context-pack designer. Build the minimum context package needed for an AI assistant to perform the task well without guessing.

TASK GOAL:
{{task_goal}}

SOURCE MATERIAL:
{{source_material}}

CONSTRAINTS:
{{constraints}}

OUTPUT REQUIREMENTS:
{{output_requirements}}

WORKFLOW SURFACE:
{{workflow_surface}}

Return exactly:
1) Essential context to include (ranked list)
2) Missing context that should be gathered before execution
3) Do-not-assume list
4) Recommended order for presenting the material to the model
5) Final context pack
6) Prompt skeleton that should be used with that pack

Rules:
- Prefer the smallest useful pack over giant dumps of material.
- Separate confirmed facts from likely assumptions.
- If source material is noisy, summarize what to keep and what to omit.
- Adapt the pack to the named workflow surface instead of assuming every tool can use the same context volume equally well.

When to Use

Use this when the task is real but the model keeps missing nuance because the source material is incomplete, messy, or buried in too much unrelated detail. It is particularly useful before analysis, writing, research synthesis, and planning tasks.

Great fits:

• long meeting notes that need a shorter usable brief
• mixed screenshots, notes, and links that need one clean package
• complex prompts where the source material is more important than the wording
• handoffs between source-grounded research, chat, editor, and coding-agent workflows

Variables

Variable	Description	Good input examples
task_goal	What the model should ultimately do	summarize, critique, plan, draft, classify, prioritize
source_material	The raw material currently available	notes, docs, code excerpt, transcript, links, logs
constraints	Limits and boundaries that must be respected	no invented facts, privacy limits, word count, no code changes
output_requirements	The desired shape of the final answer	table, memo, checklist, JSON, short brief
workflow_surface	Where the pack will be used next	chat tool, coding agent, editor assistant, internal handoff

Tips & Variations

• Add “mark missing high-risk context first” when mistakes would be expensive.
• Use this before large prompts when the problem is not wording but overloaded source material.
• For source-heavy research tasks, stage raw files in NotebookLM first, then feed the cited brief into the next tool.
• For coding work, include the likely files, failing output, and verification commands in the pack instead of pasting the whole repo.
• For research or writing tasks, ask the pack builder to separate must-read sources from nice-to-have background.

Example Output

Essential context: target audience, approved source notes, decision deadline, and known constraints.

Do-not-assume: budgets, dates, and stakeholder approval that are not explicitly present in the source material.

Prompt skeleton: a task block plus the trimmed context pack, explicit constraints, and the requested output shape.

Related site references

• Guide: Context, Constraints, and Examples
• Guide: From Chat to Code
• Tool: NotebookLM