Why
Portable authoring accepts an input_mapping source whose node does not exist or is not in the consumer's dependency closure. The scheduler can launch producer and consumer together, serializing the required input out of Context because its value is undefined. DAG-A05 reproduced both an unknown producer and an unordered producer.
Scope
Input mapping reference parsing, structural/authoring and replan validation, missing-value diagnostics, and targeted tests. Preserve documented mapping syntax and legitimate upstream output reuse; do not silently add unexpected scheduling edges.
Approach
Use GPT-5.6-Sol on this independently owned validation issue in parallel with the atomic replan foundation. Validate the referenced node and its transitive dependency relationship for ordinary start, templates and replans. Make missing required source values fail visibly at the execution boundary. Coordinate any shared runtime edits with the integrator.
This issue is one independently verifiable repair in the seven-finding DAG audit delivery. The authorized route is SpecGit delivery to dev, a separately tracked promotion to main, then the official release workflow. The integrator reviews each issue patch and records exact-head local and CI evidence before acceptance. Unrelated workspace modifications and deferred database maintenance #531 remain outside this delivery.
Acceptance
Unknown, self or unordered producer references are rejected before execution with the consumer/mapping/source identified. Valid direct and transitive upstream mappings continue to work, including supported replan reuse. Required missing outputs do not silently disappear from execution Context. Authoring, mapping, template and DAG core tests pass.
The audit baseline is cf5fd26; the DAG runtime tree is identical on dev at 711e7b5. Regression tests must assert corrected public behavior using real DAG state and deterministic boundary stubs.
Why
Portable authoring accepts an input_mapping source whose node does not exist or is not in the consumer's dependency closure. The scheduler can launch producer and consumer together, serializing the required input out of Context because its value is undefined. DAG-A05 reproduced both an unknown producer and an unordered producer.
Scope
Input mapping reference parsing, structural/authoring and replan validation, missing-value diagnostics, and targeted tests. Preserve documented mapping syntax and legitimate upstream output reuse; do not silently add unexpected scheduling edges.
Approach
Use GPT-5.6-Sol on this independently owned validation issue in parallel with the atomic replan foundation. Validate the referenced node and its transitive dependency relationship for ordinary start, templates and replans. Make missing required source values fail visibly at the execution boundary. Coordinate any shared runtime edits with the integrator.
This issue is one independently verifiable repair in the seven-finding DAG audit delivery. The authorized route is SpecGit delivery to dev, a separately tracked promotion to main, then the official release workflow. The integrator reviews each issue patch and records exact-head local and CI evidence before acceptance. Unrelated workspace modifications and deferred database maintenance #531 remain outside this delivery.
Acceptance
Unknown, self or unordered producer references are rejected before execution with the consumer/mapping/source identified. Valid direct and transitive upstream mappings continue to work, including supported replan reuse. Required missing outputs do not silently disappear from execution Context. Authoring, mapping, template and DAG core tests pass.
The audit baseline is cf5fd26; the DAG runtime tree is identical on dev at 711e7b5. Regression tests must assert corrected public behavior using real DAG state and deterministic boundary stubs.