aoa-dry-run-first
Intent
Use this skill to force a preview-first mindset before executing changes that may be destructive, irreversible, or operationally meaningful.
Trigger boundary
Use this skill when:
- the task can be simulated, previewed, or inspected before real execution
- the action may delete, restore, migrate, reconfigure, or otherwise alter a live surface
- the cost of a mistaken execution is meaningfully higher than the cost of a preview step
Do not use this skill when:
- no meaningful dry-run or preview path exists and the task is already clearly bounded and harmless
- the task is purely analytical and does not approach execution at all
- the central question is whether execution is authorized at all rather than how to preview it; use
aoa-approval-gate-check
- the task is only about preparing a sanitized artifact for sharing; use
aoa-sanitized-share
Inputs
- requested action
- available preview or dry-run mechanisms
- touched surfaces
- known limits of the preview path
Outputs
- dry-run or preview recommendation
- bounded preview result or plan
- note on what the preview does and does not prove
- recommendation for next step
Procedure
- identify whether a preview, simulation, inspect-only, or dry-run path exists
- prefer the safest bounded preview path before real execution
- make explicit what the preview covers and what it cannot guarantee
- avoid treating dry-run output as proof of total safety
- recommend the next step only after the preview has been interpreted
Contracts
- dry-run should reduce uncertainty before execution
- preview results should not be overstated
- lack of a dry-run path should be named as a risk, not hidden
- real execution should not be smuggled into a preview step
Risks and anti-patterns
- treating a preview as complete validation
- skipping preview because the real change looks small
- using a fake or weak preview path that does not touch the real risk surface
- blurring inspect-only and execute behavior
Verification
- confirm a preview or dry-run path was considered first
- confirm the preview scope was named honestly
- confirm unresolved risk after preview was not hidden
- confirm the preview step did not silently perform the real action
- confirm the recommended next step matches the preview confidence
Technique traceability
Manifest-backed techniques:
- AOA-T-PENDING-DRY-RUN-FIRST from
8Dionysus/aoa-techniques at TBD using path TBD and sections: Intent, When to use, Inputs, Outputs, Core procedure, Contracts, Risks, Validation
- AOA-T-0004 from
8Dionysus/aoa-techniques at fe7df9aba3937435b431489ed0f9c5d52690a37c using path techniques/agent-workflows/intent-plan-dry-run-contract-chain/TECHNIQUE.md and sections: Intent, When to use, Outputs, Core procedure, Validation
Adaptation points
Future project overlays may add:
- local preview commands
- dry-run limitations
- restore or recovery expectations
- project-specific inspect-only patterns
1---2name: aoa-dry-run-first3description: aoa-dry-run-first4---56# aoa-dry-run-first78## Intent910Use this skill to force a preview-first mindset before executing changes that may be destructive, irreversible, or operationally meaningful.1112## Trigger boundary1314Use this skill when:15- the task can be simulated, previewed, or inspected before real execution16- the action may delete, restore, migrate, reconfigure, or otherwise alter a live surface17- the cost of a mistaken execution is meaningfully higher than the cost of a preview step1819Do not use this skill when:20- no meaningful dry-run or preview path exists and the task is already clearly bounded and harmless21- the task is purely analytical and does not approach execution at all22- the central question is whether execution is authorized at all rather than how to preview it; use `aoa-approval-gate-check`23- the task is only about preparing a sanitized artifact for sharing; use `aoa-sanitized-share`2425## Inputs2627- requested action28- available preview or dry-run mechanisms29- touched surfaces30- known limits of the preview path3132## Outputs3334- dry-run or preview recommendation35- bounded preview result or plan36- note on what the preview does and does not prove37- recommendation for next step3839## Procedure40411. identify whether a preview, simulation, inspect-only, or dry-run path exists422. prefer the safest bounded preview path before real execution433. make explicit what the preview covers and what it cannot guarantee444. avoid treating dry-run output as proof of total safety455. recommend the next step only after the preview has been interpreted4647## Contracts4849- dry-run should reduce uncertainty before execution50- preview results should not be overstated51- lack of a dry-run path should be named as a risk, not hidden52- real execution should not be smuggled into a preview step5354## Risks and anti-patterns5556- treating a preview as complete validation57- skipping preview because the real change looks small58- using a fake or weak preview path that does not touch the real risk surface59- blurring inspect-only and execute behavior6061## Verification6263- confirm a preview or dry-run path was considered first64- confirm the preview scope was named honestly65- confirm unresolved risk after preview was not hidden66- confirm the preview step did not silently perform the real action67- confirm the recommended next step matches the preview confidence6869## Technique traceability7071Manifest-backed techniques:72- AOA-T-PENDING-DRY-RUN-FIRST from `8Dionysus/aoa-techniques` at `TBD` using path `TBD` and sections: Intent, When to use, Inputs, Outputs, Core procedure, Contracts, Risks, Validation73- AOA-T-0004 from `8Dionysus/aoa-techniques` at `fe7df9aba3937435b431489ed0f9c5d52690a37c` using path `techniques/agent-workflows/intent-plan-dry-run-contract-chain/TECHNIQUE.md` and sections: Intent, When to use, Outputs, Core procedure, Validation7475## Adaptation points7677Future project overlays may add:78- local preview commands79- dry-run limitations80- restore or recovery expectations81- project-specific inspect-only patterns