Dryrun
Uncover problems that planning alone cannot reveal — through a minimal-cost 'light Act.' Works for code and dialog alike.
Effect
Verify “will this actually work?” at minimum cost during the planning phase. Expose the gap between the design in your head and reality — before committing to a full implementation.
Planning alone has blind spots. But building the whole thing only to discover “it doesn’t work” is far too costly. Dryrun strikes the balance.
Two Modes
Code Dryrun (Classic)
Write minimal, throwaway code to validate a technical assumption. “Does this API behave as expected?” “Will the container actually start with this configuration?”
Dialog Dryrun (New)
Simulate inter-Agent dialog flows and collaboration designs to find gaps in role assignments. Without writing a single line of code, ask: “If I delegate this to the Agent, what comes back?”
When to Use
Code Dryrun:
- Before using an unfamiliar technology or API
- Before implementing something with complex dependencies
- When “it should probably work” lacks conviction
Dialog Dryrun:
- Designing a new Agent or subordinate
- Designing inter-Agent collaboration flows
- When “will this Agent actually do what I ask?” is unclear
- Hunting for improvement points in existing Agents
Both:
- At a fork in the plan (debating which approach to take)
How It Works
1. Narrow Down to One Thing to Verify
Code: “Does this API return the expected response?” Dialog: “If I ask this Agent to review, will it ask clarifying questions when needed?“
2. Try It at Minimum Cost
Code: Throwaway verification code. Perfection not required. Dialog: Mentally simulate the actual scenario. “Eris → harness: request this → harness does that → gap found here.”
3. Feed the Result into Your Decision
- It worked / flowed → proceed to full implementation/operation (with confidence)
- Problem found → identify the root cause. Fix the design before moving forward
- Unexpected discovery → fold it back into the plan (this is where the real value lies)
Combinations
| Card | Synergy |
|---|---|
| Dev-Flow | A decision point exists between planning and implementation to insert a Dryrun |
| Devil / Devil Gate | Validate Dryrun results with Devil → higher quality in the real implementation |
| Reflect | Accumulate “unexpected discoveries” from Dryrun into the knowledge DB |
| Ctx | Record Dialog Dryrun results in ctx → continue improving in the next session |
”…Thinking alone won’t show you everything. Move your hands — but minimally. Imagine the dialog — but concretely.” — Eris
Want to use this card permanently? Convert it directly into a SKILL.md — Effect → Overview, When to Use → Trigger, How It Works → Steps.