fix(completion-integrity): carry final verdict into learning

[JustinJLeopard]

Summary

• pass the synthesized final run status into trajectory learning
• permit only the defined conservative complete-to-partial downgrade
• add one end-to-end regression for fidelity-downgraded all-done results

Verification

• RED: exact regression stored success while synthesis returned partial
• Python 3.14: 468 passed, 14 subtests passed
• Python 3.12: 468 passed, 14 subtests passed
• git diff --check

Source integration only. No installation, runtime, deployment, acceptance, governance, or canonical state is changed.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
justai-demo	Ready	Preview	Aug 9, 2026 8:59pm

[Copilot]

Pull request overview

This PR fixes a completion-integrity gap by ensuring the learning/trajectory layer records the synthesized final run verdict (which may be downgraded by post-execution gates like intent fidelity), rather than re-deriving a potentially incorrect “success” from raw task tallies.

Changes:

• Pass summary.status from the synthesizer into learning.record_run(..., final_status=...) so learning reflects the final verdict.
• Enforce a conservative invariant in learning: only allow the defined downgrade transition (complete → partial) when final_status differs from the tallied status.
• Add an end-to-end regression test verifying that a fidelity-downgraded run is stored as outcome="partial".

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
tests/test_orchestrator_pipeline.py	Adds regression coverage ensuring learning stores the synthesized (downgraded) outcome.
justai/orchestrator.py	Threads the synthesizer’s final summary.status into learning via record_run(..., final_status=...).
justai/learning.py	Accepts and validates final_status (only permitting complete → partial divergence) before mapping to stored outcome.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.