Repository navigation
Qualify native recovery and a pinned local GPU workflow - #6
Merged
Merged
Conversation
seathatflowsinourveins
added a commit
that referenced
this pull request
Sep 27, 2026
… are an allowed dispatch mode (#388) Practice gap #6: tools/adoption/install_skills.py now says what it checked. The SKILL.md sha256 matches the pin, and the lock records the tree. Practice gap #8: agent teams are an allowed dispatch mode, per the current Claude Code docs (agent-teams, context and communication). - examples/claude-native/CLAUDE.md gains one word-neutral sentence (924 words). - scripts/adoption_status.py reports agent_teams_opt_in (0 or 1) as information, not agent_teams_off as a wiring requirement. - docs/token-efficiency-stack.md is updated to match. - docs/harness-rules-convergence-20260922.md marks WH-3 superseded. GPT-6 cross-family review: accept, no findings (213 tests, OK). After main (#389, #387) was merged in: validate.py passed; tests.test_adoption_status and tests.test_landscape_sweep_harness OK; CI full suite green. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012qwXNtFyUhY5x2dtrm7tG5
This was referenced Sep 27, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Runtime checkpoint retry and native session recovery were unqualified, and the clean WSL runtime lacked a tested local inference lane. This change publishes frozen, scoped acceptance for Mac/WSL Dagu retry, Mac Codex same-session recovery, WSL Claude same-session recovery, and one pinned Qwen3.8-27B partial-offload repair. Checkpoints execute once, unfinished work completes, and native usage retains its original accounting boundaries.
Integrates the reviewed token-practice branch from #4 while preserving current convergence and workflow-security checks. The explorer now links to the integrated source revision; component and evidence references are reciprocal. The local GPU remains optional: its one accepted generation took 179.191 seconds, exact layer count was unavailable, and no frontier-parity or savings claim is made. No accounts, private transcripts, model weights or machine paths are published.
Validation: 429 offline tests passed with 47 explicit environment-dependent skips on Mac, including a nested 11-test Claude guard suite. All five new convergence records, catalog joins, integrity checks, explorer rebuild and secret scan passed. Separate native fixtures and retained failed attempts are linked in the receipts. Private-log placement and cleanup failure handling were hardened after review without repeating model trials.