fix(sleep): harden merged outcome, grouping, and Copilot paths

[Yifan Yang (Yif-Yang)]

Follow-up hardening for #198, #186, and #200 after preserving their original merge commits and contributor attribution.

Changes

• Measure bare-refusal length after removing leading Markdown markers, while avoiding Unicode case-folding length distortion.
• Enforce read-only shared memory in grouped consolidation without duplicate evolve_memory keyword failures.
• Redact Copilot CLI transcript secrets before the 4000-character persistence boundary is applied.

Verification

• 86 focused tests passed.
• Full suite: 796 passed, 6 skipped, 130 subtests passed.
• Ruff passed on all changed source and test files.
• git diff --check passed.
• A second independent Codex review found no blockers.

[Copilot]

Pull request overview

This PR hardens SkillOpt-Sleep’s consolidation and judging/harvest pipelines following earlier merges, focusing on refusal detection robustness, enforcing read-only shared memory in grouped consolidation, and preventing secret persistence when harvesting Copilot CLI sessions.

Changes:

• Update no_refusal detection to measure refusal length after stripping leading Markdown markers and avoid Unicode lowercasing length distortion.
• Enforce shared-memory isolation in consolidate_groups() by overriding any caller-supplied evolve_memory.
• Redact secrets in Copilot CLI session text before applying the 4000-character clipping bound.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/test_sleep_multi_skill.py	Adds coverage ensuring callers cannot override grouped shared-memory isolation.
tests/test_outcome_judges.py	Adds regression tests for refusal-length measurement vs Markdown prefixes and Unicode lowercasing expansion.
tests/test_harvest_copilot_cli.py	Adds tests asserting secrets are redacted (including before clipping) when harvesting Copilot CLI sessions.
skillopt_sleep/multi_skill.py	Hardens grouped consolidation to always force evolve_memory=False without duplicate-kw failures.
skillopt_sleep/judges.py	Refines refusal detection to strip Markdown markers before length checks and avoid lowercasing length distortion.
skillopt_sleep/harvest_copilot_cli.py	Redacts secrets before truncating harvested Copilot CLI text.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.