Make money doing the work you believe in
In 2026, everyone in AI is talking about Context Engineering.
It's already the core AI PM skill.
But just a few explain what the context is.
Save this template. It captures all scenarios and will help you maximize agents' performance:
1. Instructions
Define:
→ Who: Encourage an LLM to act as a persona
→ Why is it important (motivation, larger goal, business value)
→ What are we trying to achieve (desired outcomes, deliverables, success criteria)
If you remember one thing from this post, remember this: 𝗱𝗼𝗻’𝘁 𝗱𝗲𝗳𝗮𝘂𝗹𝘁 𝘁𝗼 𝗰𝗼𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗽𝗿𝗼𝗺𝗽𝘁𝘀. Providing strategic context helps AI understand your intent and make better decisions (arXiv:2401.04729)
2. Requirements (How)
Define:
→ Steps to take (reasoning, tasks, actions)
→ Conventions (style/tone, coding rules, system-design)
→ Constraints (performance, security, test coverage, regulatory)
→ Response format (JSON, XML, plain text)
→ Examples (positive/negative, responses/behaviors)
Negative examples might help you address issues identified during error analysis
3. Knowledge
Define:
→ External Context:
- Domain (strategy, business model, market facts)
- System (overall goals, other agents/services)
→ Task Context:
- Workflow (process steps, process, hand‑offs)
- Documents (specs, procedures, tickets, logs)
- Structured Data (variables, tables, arrays, JSON/XML)
4. Memory
An LLM can access:
→ Short-term memory
- Previous messages, chat history
- State (e.g., reasoning steps, progress)
→ Long-term memory
- Semantic (facts, preferences, user knowledge)
- Episodic (experiences, past interactions)
- Procedural (instructions from previous interactions)
Memory is not part of the prompt you can type. It can be automatically attached by the orchestration layer or accessed as a tool.
5. Tools
Provide description, what it does, how to use it, return value, parameters.
It's special “functions” block in the LLM context window. It does consume your input tokens and affect the performance.
Treat tool descriptions as micro-prompts that guide agents’ reasoning.
Descriptions provided by MCP servers are often insufficient and do not consider your specific domain context.
6. Tool results
Key to understand:
To call a function, an LLM uses a special format interpreted by the system. It’s like saying, “Please call this tool with these parameters."
Next, an orchestration layer responds by attaching a special message to the messages list.
—
Free examples (GitHub):
PM Agent Context: github.com/phuryn/examp…
Bug Fixing Agent Context: github.com/phuryn/examp…
—
Find this helpful?
You might also like this Ultimate Guide to Context Engineering: productcompass.pm/p/con…
Want to pivot to AI PM?
I recommend the AI PM Certification. It’s a 6-week cohort by Miqdad Jaffer (Product Lead at OpenAI). I lead AI Builds Labs. The next session: Oct 18, 2025. Next cohort: January 26. $500 off: bit.ly/aipmcohort
