LLM-Friendly Context

The goal is stable downstream execution: the next agent should know what to read, what to do, what counts as success, and when to stop or escalate.

An active workflow's declared input contract is already optimized for its specialist. For that handoff, preserve the contract's named fields and declared value forms; the prompt consists of those fields and values. Apply the general prompt-composition rules below only when no input contract exists, and apply the generated-artifact rules to artifacts.

Core Rules

1. Use positive, executable instructions

   • State what the next agent should do.
   • Convert quality policies into positive criteria.
   • Keep a prohibition only when it protects an irreversible boundary or shipped contract. Name the protected condition and the allowed action.
   • Example: "Preserve existing public API behavior across the documented compatibility cases."

2. Make vague instructions concrete

   • Replace subjective terms with observable conditions, paths, commands, schemas, examples, or decision rules.
   • Terms that often need clarification when they leave a decision to the next agent: appropriate, proper, related, existing behavior, optional, as needed, if needed, per convention, unresolved alternatives, TBD, placeholder.

3. Specify output shape

   • Define required sections, fields, table columns, JSON keys, or checklist items.
   • For handoffs, include paths to produced artifacts and the exact status fields the caller must inspect.

4. Provide necessary context

   • Include the purpose, source artifacts, hard constraints, accepted decisions, and unresolved conditions.
   • Prefer concrete file paths and section hints over broad module names.
   • Follow references while they can change an in-scope decision, action, or verification result.

5. Decompose complex work into verifiable steps

   • Split work with 3+ objectives or sequential dependencies into ordered steps.
   • Each step needs a checkpoint: what evidence proves it is complete.

6. Permit uncertainty explicitly

   • If the source material is missing, contradictory, or not verifiable, state the uncertainty and the required escalation.
   • Record unknown business, product, security, or compatibility decisions as blocking unresolved items with the input needed to resolve them.

7. Keep constraints proportionate

   • Add only constraints that reduce ambiguity or preserve a real requirement.
   • Keep simple downstream tasks lightweight when the target action, context, and success criteria are already clear.
   • Apply minimal, a few lines, and explicit line estimates to the completed diff as one total budget.

Rewrite Patterns

Use these rewrites before treating a prompt, handoff, or artifact as complete.

Handoff Checklist

Before sending a prompt or artifact to another agent, verify:

• The target action is explicit.
• Required input paths and source artifacts are named.
• Accepted decisions and constraints are stated once, without alternate wording.
• Output format or expected status fields are specified.
• Success criteria are observable.
• Ambiguous expressions have been rewritten or marked as unresolved.
• Each instruction states the allowed action; each retained prohibition names the protected condition and allowed alternative.
• The next agent can complete its scope with explicit choices, decision rules, or blocking unresolved items.

Generated Artifact Checklist

Before writing or finalizing a generated document:

• Each requirement, claim, task, test skeleton, or review finding has enough source context to trace why it exists.
• Every executable instruction names the target, action, and expected result.
• Verification steps say what to run or observe and what result proves success.
• Each instruction states the allowed action; each retained prohibition names the protected condition and allowed alternative.
• If an artifact is derived from another artifact, copied decisions stay consistent in wording and meaning.
• If downstream work is blocked by missing information, the artifact records the missing input and escalation condition.