Agent Brief

Purpose

What project improvement should this run make?

Aim

What outcome or behavior change does this serve?

Problem statement

What problem framing did we select?

Selected solution level

Choose one:

• [ ] Band-Aid
• [ ] Local Optimum
• [ ] Reframe
• [ ] Redesign

Why this level?

Why not the other levels?

• Band-Aid:
• Local Optimum:
• Reframe:
• Redesign:

Small strategy

Mechanism: Why should this approach move the aim?

Feedback: What signal will show quickly whether it worked?

Guardrails: What must not break or drift?

- - -

Context

The agent should inspect:

- - -

Relevant constraints:

- - -

Prior attempts, symptoms, or review comments:

- - -

Behavior contract

The change must:

- - -

The change must not:

- - -

Checks

Commands to run:

Acceptance checks:

- - -

Old behavior these checks should catch:

- - -

Stop conditions

The agent should stop and report instead of guessing if:

• the selected problem framing is wrong;
• required files or commands are missing;
• the patch requires a broader redesign than this slice allows;
• tests cannot be made meaningful without changing the brief;
• implementation would touch unrelated systems.

Review checklist

Before accepting output, check:

• [ ] Does it serve the aim?
• [ ] Did it stay at the selected solution level?
• [ ] Is the mechanism visible in the code or tests?
• [ ] Do checks fail on the old behavior?
• [ ] Did the agent run the right commands?
• [ ] Are edge cases named and tested?
• [ ] Did it avoid parallel paths or compatibility shims?
• [ ] Are unrelated changes absent?
• [ ] Is the result easier for the next maintainer to edit?