Installs into .claude/skills of the current project.
Are you the author of Trace Assumption Cascade?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/yogsoth-ai-trace-assumption-cascade)
---
name: trace-assumption-cascade
description: "Trace downstream claims, methods, and conclusions that fail when a selected assumption is invalidated; distinguish local from cascading failure."
---
# trace-assumption-cascade
## Purpose
Trace downstream claims, methods, and conclusions that fail when a selected assumption is invalidated.
## Input contract
```yaml
required: [assumption, dependency_graph, downstream_claims]
optional: [failure_evidence, intervention]
constraints: [each cascade edge must identify dependency and failure mechanism]
```
## Procedure
1. Identify the assumption and immediate dependents.
2. Propagate invalidation through claims, methods, and conclusions.
3. Distinguish local from cascading failure and record recovery points.
If the cascade identifies conclusions controlled by a small assumption set, consider `search-minimal-flip` as the next tactic.
## Output contract
```yaml
produces: [assumption_cascade, affected_claims, local_failures, cascading_failures, recovery_points]
delta_fields: [findings, evidence_updates, uncertainties, decisions, open_questions]
```
## Quality gates
- A-class cascade coverage: declared universe = all downstream nodes reachable from the selected assumption; numerator = nodes with an assessed dependency and failure status; batch increment = one dependency layer; stopping reason = no new reachable nodes or marginal new impact is below a justified threshold; source references = dependency edges, assumption IDs, evidence IDs; direction/threshold reason = continue while newly reached nodes materially change decision or risk.
## Failure and counterexamples
Do not infer cascade from naming or proximity; require a dependency edge.
## Provenance map
- resolved: assumption-cascade