eda-theory-veto
Theory-level veto skill for EDA plans. Use before expensive experiment submissions or major flow/model changes to identify logically unsound assumptions, contradiction with known physics/policy, and high-risk invalid comparisons. Produces GO/CONDITIONAL/NO-GO with evidence.
EDA Theory Veto
Overview
This skill protects the loop from executing invalid ideas just because they are executable. It evaluates proposal validity against theory, known constraints, and policy, then emits a veto decision.
When to run
Run before:
- launching large/long job batches,
- changing objective/cost model assumptions,
- making algorithmic comparison claims,
- introducing new route/CTS/PDN policy that may break prior assumptions.
Inputs
Required:
- current user proposal (one-paragraph objective),
- relevant artifacts (latest summary/gate/monitor),
- KB references used by the proposal.
Recommended:
- primary-source/web references when assumptions may drift.
Workflow
- Parse proposal into falsifiable assumptions.
- Check each assumption against:
- known physics/EDA constraints,
- local policy locks (comparison, route policy),
- current empirical evidence.
- Classify conflicts:
- hard contradiction,
- missing evidence,
- plausible but high-risk.
- Emit one verdict:
GO: no blocking contradiction.CONDITIONAL: proceed only with listed constraints.NO-GO: block execution; provide minimum fix path.
- Provide minimal corrective experiment if vetoed.
Decision output contract
Always output:
verdict: GO / CONDITIONAL / NO-GOblocking_reasons: numbered listassumption_table: assumption, evidence, statussafe_next_step: one minimal testable actionoverride_policy:- NO-GO can be bypassed only with explicit user override.
- when override happens, mark run as
veto_overridden.
References
Load:
references/veto-criteria.mdreferences/decision-template.md