Week 2 Reflection: First Run, First Judgment¶
Purpose¶
This week, you did two important things:
1. You used OpenCode to complete a small task.
2. You used /init to create project instructions in AGENTS.md.
The goal of this reflection is not to prove that the agent was perfect. The goal is to notice how you worked with it.
Good agent users do not blindly accept output. They ask better questions, check evidence, and keep responsibility for the final result.
Estimated time: 15-20 minutes.
Part A: What Happened in Lab 2.1?¶
Think back to the typo-fix task.
Answer in 2-4 sentences each:
- What did you ask OpenCode to do first?
- Did the agent inspect the correct file? How could you tell?
- What changed when you switched from plan mode to build mode?
- Did the final file contain the expected fix?
- What did you personally verify instead of trusting the agent's summary?
Part B: What Happened in Lab 2.2?¶
Think back to the /init and AGENTS.md lab.
Answer in 2-4 sentences each:
- What project did you use for
/init? - What did
AGENTS.mdcorrectly understand about the project? - What part of
AGENTS.mdwas vague, incomplete, or possibly wrong? - What evidence did the agent use from the repo?
- What did you understand better after reading
AGENTS.md?
Part C: When Did the Agent Get It Wrong?¶
Every agent makes mistakes. A mistake can be small: - It guesses the wrong test command. - It describes a folder too generally. - It misses an important convention from the README. - It sounds certain about something that is only a guess.
Choose one moment from this week when the agent was wrong, incomplete, or too confident.
Use this structure:
What the agent said or did:
Why I think it was wrong or incomplete:
What clue helped me notice:
What I would ask next time:
If you did not catch any mistake, choose the weakest or vaguest part of the agent's output instead.
Part D: Your Current Mental Model¶
Complete these sentences:
- Plan mode is useful when I want to...
- Build mode is useful when I want to...
/initis useful because...AGENTS.mdshould not be trusted blindly because...- Before I let an agent edit files, I should...
Part E: Confidence Check¶
Rate each statement from 1 to 5.
| Statement | Rating |
|---|---|
| I can start OpenCode from inside a project folder. | 1 2 3 4 5 |
| I can tell whether I am in plan mode or build mode. | 1 2 3 4 5 |
I can run /init and find the generated AGENTS.md. |
1 2 3 4 5 |
| I can explain project conventions in my own words. | 1 2 3 4 5 |
| I can notice when the agent may be guessing. | 1 2 3 4 5 |
If any rating is 1 or 2, write one sentence about what you need to practice before Week 3.
Part F: One Rule for Your Future Self¶
Write one personal rule you want to follow when using OpenCode.
Examples:
- "I will use plan mode before build mode on unfamiliar projects."
- "I will ask the agent what evidence supports its conclusion."
- "I will read AGENTS.md before asking for edits."
- "I will verify changes myself before considering a task done."
Your rule:
Completion Checklist¶
You are done with Week 2 when:
1. OpenCode is installed and working.
2. Your model provider is configured.
3. You completed the typo-fix task from Lab 2.1.
4. You ran /init in a real project from Lab 2.2.
5. You read and reviewed AGENTS.md.
6. You can explain one thing the agent got right and one thing it got wrong or left vague.
Looking Ahead¶
In Week 3, you will practice the habit that separates safe agent use from risky agent use:
Explore first. Edit second.
You will use plan mode to investigate a project before making changes. The Week 2 skill you need most is simple: pause long enough to understand what the agent thinks it knows.