Root-Cause Claim Discipline
This release makes a structural fix to the most common debugging failure pattern: stopping at an intermediate breakpoint or ripple and presenting it as the root cause. It does so by turning the stop condition from a self-judged claim into a set of mechanical, falsifiable checks, and by replacing the implicit single-root default with explicit causal topology classification.
This is method-pack advisory discipline. It is not a GateDecision, PolicySnapshot, or completion authority (per ADR-0001).
Pre-Claim Gate (systematic-debugging Phase 3.5)
When a patch-shape signal fires — the candidate fix is a guard, fallback, consumer/caller patch, artifact/cache patch, or sample-only naming — the agent must pass five mechanical checks before claiming a root cause or entering the fix phase:
- Causal Closure — every causal edge from symptom to claimed root has an evidence anchor; one "probably" edge leaves the chain open.
- Falsifier Checked — state "if X were not the root cause, observable F would appear," and confirm F was checked and absent.
- Adversarial Self-Refutation — generate the strongest single argument that this root cause is wrong, and show why it does not hold.
- Causal Topology Gate — classify the topology explicitly (see below).
- Layer Ceiling Proof — if the claim stops at L?, show why L?+1 is unreachable by concrete constraint, not by omission.
The quick bug lane is exempt when no patch-shape signal fires, so fast-path cheapness is preserved.
Causal Topology Gate
Replaces the implicit "single-root" default with explicit classification into six causal topologies. The default is unknown; the agent must actively exclude the multi-root topologies before collapsing to a single-root claim.
Boundary
This is advisory method-pack discipline. A method pack cannot enforce the gate at runtime; it can only make a gate-miss visible so a human partner can catch it. Reducing the failure rate to zero requires runtime enforcement, which ADR-0001 reserves for a future independent runtime core.
Known limitations
- Necessity tests for large clusters are conceptual proofs, not empirical test runs — a method-pack ceiling.
- Cyclic causation (A → B → A) is not covered by the six topologies; the doc flags this for extension if a domain has feedback loops.
Verification
All automated checks pass:
tests/e2e/layer1-fast-check.sh --host-profile none— 25/25tests/e2e/boundary-compliance-check.sh— pass (advisory wording not flagged)tests/e2e/debugging-patch-shape-gate-check.sh— 35 assertionstests/e2e/workflow-quality-check.sh— passtests/helpers/validate_workflow_quality_matrix.py— pass