[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78SRRF5CX4VABGA2F4EF8T0":3,"skill-same:01M78SRRF5CX4VABGA2F4EF8T0":60},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M78SRRF5CX4VABGA2F4EF8T0","melee-pr-workflow",null,"Top-level Melee PR workflow skill. Use when syncing the orchestrator-owned past-PR corpus, syncing or rebasing the parent doldecomp\u002Fmelee checkout, designing reviewer-friendly PRs or PR series, preparing a PR for handoff, running PR regression checks, drafting\u002Fupdating PR bodies, or coordinating GitHub PR review\u002FCI work.","fjooord\u002Fgamecube-decomp-harness","fjooord","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F44952366?v=4","01M75A80KS3BN58VMNBG7AXAM1","Ford Lascari","01M75A80MCVXVSXGREY7Y9Z8H5",7,"MIT","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fmelee-pr-workflow\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: melee-pr-workflow\ndescription: \"Top-level Melee PR workflow skill. Use when syncing the orchestrator-owned past-PR corpus, syncing or rebasing the parent doldecomp\u002Fmelee checkout, designing reviewer-friendly PRs or PR series, preparing a PR for handoff, running PR regression checks, drafting\u002Fupdating PR bodies, or coordinating GitHub PR review\u002FCI work.\"\n---\n\n# Melee PR Workflow\n\nThis is the single top-level skill for Melee PR-adjacent work. It routes to the\ndecomp orchestrator for missing PR knowledge sync and regression gates, while keeping\nparent Melee git\u002FPR work separate from the nested `decomp-orchestrator\u002F` repo.\n\n## Guardrails\n\n- Treat `decomp-orchestrator\u002F` as an unrelated nested Git repository unless the user explicitly asks to work on the orchestrator itself.\n- Run parent Melee git sync\u002Frebase commands from the parent Melee checkout, not from inside `decomp-orchestrator\u002F`.\n- Mainline is `master`; use `origin\u002Fmaster` as the rebase and regression baseline.\n- Do not stage, commit, reset, rebase, or push `decomp-orchestrator\u002F` as part of parent Melee PR work.\n- Inspect `git status --short --ignore-submodules=all` before mutating the parent checkout.\n- Never use `git reset --hard` or discard local work as part of this workflow.\n\n## Refresh PR Knowledge\n\nUse this path when the user asks to sync missing PR data, comments, reviews,\ndiffs, searchable postmortems, file-card context, or past-PR graph\u002Fsearch data.\n\nPreview the fetch scope:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py --dry-run\n```\n\nFetch only PRs that are not already present locally and scaffold searchable records:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py\n```\n\nRun Pi-reviewed postmortems for PRs that need searchable records:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fbuild_pr_postmortems.py \\\n  --dump-root decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fdata \\\n  --run-agent \\\n  --jobs 16\n```\n\nRebuild only the past-PR graph slice after corpus changes:\n\n```bash\nbun run --cwd decomp-orchestrator kg:rebuild -- --repo-root \"$PWD\" --sources past_prs\n```\n\nNotes:\n\n- `kg-maintain` and `trigger-agent` can index pending postmortems and rebuild graph state, but they do not fetch fresh GitHub PR data.\n- The PR corpus lives under `decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fdata`.\n- Existing PR dump slices are treated as immutable during sync. To rebuild a\n  stale or damaged PR record, delete that PR's local slice first, then run the\n  missing-only fetcher.\n- The fetcher uses `gh api`; GitHub CLI auth must be available.\n\n## Sync Repo And PR Corpus\n\nUse this path when the user asks to sync, rebase, or update the local checkout\nand missing PR library entries together.\n\nStandard sync:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\nSkip git and only sync missing PR corpus entries:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --skip-git \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\nIf rebase stops, leave the worktree in Git's rebase state and report the\nconflicted files plus Git's suggested next command.\n\n## Shape PRs And PR Bodies\n\nUse this path when the user asks to split, group, restructure, create, update,\nor explain Melee PRs.\n\nLoad `references\u002Fpr-shaping-reviewer-guidance.md` before proposing or applying a\nPR split. Design PRs around reviewer cognition and regression-tool signal:\n\n- **Only exact matches ship.** Match PRs carry runner-validated exact matches\n  (plus supporting headers\u002Fdeclarations the matches need to build); they are\n  byte-verified and should read as easy approvals. Fuzzy improvements never go\n  into PRs — they stay on the local branch (the split plan's `local-*` slices)\n  until they become matches. If a match-lane file also carries unshipped fuzzy\n  improvements in other functions, call that out in the PR body.\n- Treat the configured max-files-per-PR as a hard ceiling, not a packing\n  target. Aim for the fewest PRs that are still comfortable to review in one\n  sitting; do not shave slices down to produce a pile of small PRs.\n- Keep broad header, naming, symbol, build, metadata, and rename churn separate\n  from ordinary implementation\u002Fmatching work when possible.\n- Group small changes only when they share one review context and risk class.\n- Treat shared headers, symbol renames, and bot-visible name changes as\n  high-risk even when the file count is small.\n- Avoid path-only splits when a subsystem, actor\u002Fmodule family, matching\n  milestone, API\u002Fheader change, or mechanical cleanup category gives reviewers\n  a clearer unit.\n- Write PR bodies as reviewer-facing digests using the reference template:\n  summary, PR shape, reviewer notes, and verification\u002Fregression status.\n\n## Prepare PR Handoff\n\nUse this path when the user asks to prepare, finalize, refresh, build-check,\nregression-check, update, or hand off a Melee PR.\n\nThe dashboard's `Prepare Handoff` automates the core of this path (pause →\npull upstream & rebase → PR intake → rebuild the production baseline in a\nper-SHA worktree → branch QA vs that baseline → checkpoint with regressed\nsymbols forced to needs_rework → requeue rework at repair priority →\nmatch-only split plan → ship-set verification, where the match-lane diff is\napplied onto the baseline worktree and must build with zero regressions) and\nshares the `\u002Ftmp\u002Fmelee-baseline-\u003Csha>` cache with step 5 below. Use the\nmanual steps for PR-comment triage, PR-body work, and pushing.\n\nFor review-style cleanup and regression triage, use the local QA reference in\n`references\u002Fpr-review-qa-standards.md`. Keep that reference in the workflow\nskill, not in upstream Melee docs.\n\nFor adversarial cleanup of an existing PR, standards-clean source outranks fuzzy\nscore. If removing overzealous worker output, pragmas, fake anchors, dummy\npadding, or other tactic-shaped code lowers fuzzy score or even loses a match,\nkeep the clean source, report the score impact, and route the result through\ncarry-forward or explicit operator policy instead of restoring the tactic.\n\nFor PR splitting\u002Fgrouping and reviewer-facing PR body content, use\n`references\u002Fpr-shaping-reviewer-guidance.md`.\n\n1. Inspect status and PR context:\n\n```bash\ngit status --short --branch --ignore-submodules=all\ngit -C decomp-orchestrator status --short\ngit remote -v\ngh pr status --repo doldecomp\u002Fmelee\n```\n\n2. Sync mainline and missing PR knowledge:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\n3. Pull active PR comments and reviews:\n\n```bash\ngh pr view --repo doldecomp\u002Fmelee --comments\ngh pr view --repo doldecomp\u002Fmelee --json number,url,title,state,isDraft,baseRefOid,headRefOid,reviewDecision,mergeStateStatus\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fpulls\u002F\u003CPR_NUMBER>\u002Fcomments\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fissues\u002F\u003CPR_NUMBER>\u002Fcomments\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fpulls\u002F\u003CPR_NUMBER>\u002Freviews\n```\n\n4. Evaluate PR shape before validation:\n\n- Inventory touched file types and high-risk churn.\n- Check the latest checkpoint (`pr_candidates.md`): only the match candidates\n  ship; notable improvements are listed in `carry_forward.md` and stay local.\n- Decide whether the PR should stay grouped, split, or isolate header\u002Fnaming\n  changes.\n- Draft\u002Fupdate the PR body with a clear summary, PR shape, reviewer notes, and\n  verification section.\n- Call out expected regression false positives caused by renames or declaration\n  movement.\n\n5. Rebuild a baseline from current `origin\u002Fmaster`:\n\n```bash\nBASE_SHA=\"$(git rev-parse origin\u002Fmaster)\"\nBASE_DIR=\"\u002Ftmp\u002Fmelee-baseline-${BASE_SHA}\"\nif [ ! -f \"$BASE_DIR\u002Fbuild\u002FGALE01\u002Fbaseline.json\" ]; then\n  git worktree add --detach \"$BASE_DIR\" \"$BASE_SHA\"\n  (cd \"$BASE_DIR\" && ninja baseline)\nfi\ncp \"$BASE_DIR\u002Fbuild\u002FGALE01\u002Fbaseline.json\" build\u002FGALE01\u002Fbaseline.json\n```\n\n6. Run build and regression gates:\n\n```bash\nninja\nbun run --cwd decomp-orchestrator orch -- \\\n  --repo-root \"$PWD\" \\\n  --state-dir \"$PWD\u002F.decomp-orchestrator-state\u002Fpr-prepare\" \\\n  regression-check \\\n  --target changes_all \\\n  --report-title \"Report for GALE01 PR handoff\" \\\n  --report-max-rows 300\n```\n\n7. Fix and rerun until the handoff state is clean:\n\n- `build\u002FGALE01\u002Fmain.dol: OK`\n- zero broken matches\n- zero unexplained or unaccepted fuzzy regressions in unmatched items\n- zero unit, section, or function metric regressions\n- no unresolved actionable review comments or relevant build warnings\u002Ferrors\n\n8. Run the mandatory pre-ship adversarial review gate (blocks handoff):\n\nSave the match-only split plan as JSON if you do not already have one, then run\nthe pr-review agent in preship mode over every shipping slice:\n\n```bash\nbun run --cwd decomp-orchestrator orch -- \\\n  --project melee \\\n  pr-split-plan --checkpoint \u003Ccheckpoint.json> --ship-status \u003Cship_status.json> --json \\\n  > \u002Ftmp\u002Fmelee-pr-split-plan.json\nbun run --cwd decomp-orchestrator orch -- \\\n  --project melee \\\n  pr-preship-review --plan \u002Ftmp\u002Fmelee-pr-split-plan.json --all\n```\n\n- ANY `reject` finding blocks handoff; exit 1 means do not draft PR bodies or\n  push. Agent\u002Ftool failures also exit 1 (the gate fails closed) — fix the\n  infrastructure and rerun rather than skipping the gate.\n- Disposition of rejects, consistent with the MATCHES-only promotion policy:\n  mark the affected symbols `needs_rework`, requeue them at repair priority,\n  and ship the slice without them or not at all. Never keep a rejected hunk to\n  preserve a match score.\n- Per-slice verdicts, findings, and prompts land under\n  `decomp-orchestrator\u002Fprojects\u002Fmelee\u002Fstate\u002Fpreship_reviews\u002F\u003Crun-id>\u002F\u003Cslice-id>\u002F`\n  (`review.md` is the human-readable digest); cite them in the PR body's\n  verification section.\n\n9. Commit or amend only parent Melee PR files, then push safely:\n\n```bash\ngit status --short --untracked-files=no --ignore-submodules=all\ngit diff --check\ngit add \u003Cmelee-source-files>\ngit commit --amend --no-edit\ngit push --force-with-lease fork HEAD:\u003Cbranch>\n```\n\n10. Update or create the PR, then watch CI:\n\n```bash\ngh pr edit \u003CPR_NUMBER> --repo doldecomp\u002Fmelee --body-file \u003Creport.md>\ngh pr checks \u003CPR_NUMBER> --repo doldecomp\u002Fmelee\ngh run view \u003CRUN_ID> --repo doldecomp\u002Fmelee --log-failed\n```\n\nInclude PR URL, head SHA, base SHA, changed units, match\u002Fregression summary,\nlocal gate results, and CI state in the final response and PR body.\n\n## Verification\n\nFor skill maintenance or command sanity checks:\n\n```bash\npython3 -m py_compile \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fbuild_pr_postmortems.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fmigrate_pr_data_layout.py\n```\n","# Melee PR Workflow\n\nThis is the single top-level skill for Melee PR-adjacent work. It routes to the\ndecomp orchestrator for missing PR knowledge sync and regression gates, while keeping\nparent Melee git\u002FPR work separate from the nested `decomp-orchestrator\u002F` repo.\n\n## Guardrails\n\n- Treat `decomp-orchestrator\u002F` as an unrelated nested Git repository unless the user explicitly asks to work on the orchestrator itself.\n- Run parent Melee git sync\u002Frebase commands from the parent Melee checkout, not from inside `decomp-orchestrator\u002F`.\n- Mainline is `master`; use `origin\u002Fmaster` as the rebase and regression baseline.\n- Do not stage, commit, reset, rebase, or push `decomp-orchestrator\u002F` as part of parent Melee PR work.\n- Inspect `git status --short --ignore-submodules=all` before mutating the parent checkout.\n- Never use `git reset --hard` or discard local work as part of this workflow.\n\n## Refresh PR Knowledge\n\nUse this path when the user asks to sync missing PR data, comments, reviews,\ndiffs, searchable postmortems, file-card context, or past-PR graph\u002Fsearch data.\n\nPreview the fetch scope:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py --dry-run\n```\n\nFetch only PRs that are not already present locally and scaffold searchable records:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py\n```\n\nRun Pi-reviewed postmortems for PRs that need searchable records:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fbuild_pr_postmortems.py \\\n  --dump-root decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fdata \\\n  --run-agent \\\n  --jobs 16\n```\n\nRebuild only the past-PR graph slice after corpus changes:\n\n```bash\nbun run --cwd decomp-orchestrator kg:rebuild -- --repo-root \"$PWD\" --sources past_prs\n```\n\nNotes:\n\n- `kg-maintain` and `trigger-agent` can index pending postmortems and rebuild graph state, but they do not fetch fresh GitHub PR data.\n- The PR corpus lives under `decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fdata`.\n- Existing PR dump slices are treated as immutable during sync. To rebuild a\n  stale or damaged PR record, delete that PR's local slice first, then run the\n  missing-only fetcher.\n- The fetcher uses `gh api`; GitHub CLI auth must be available.\n\n## Sync Repo And PR Corpus\n\nUse this path when the user asks to sync, rebase, or update the local checkout\nand missing PR library entries together.\n\nStandard sync:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\nSkip git and only sync missing PR corpus entries:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --skip-git \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\nIf rebase stops, leave the worktree in Git's rebase state and report the\nconflicted files plus Git's suggested next command.\n\n## Shape PRs And PR Bodies\n\nUse this path when the user asks to split, group, restructure, create, update,\nor explain Melee PRs.\n\nLoad `references\u002Fpr-shaping-reviewer-guidance.md` before proposing or applying a\nPR split. Design PRs around reviewer cognition and regression-tool signal:\n\n- **Only exact matches ship.** Match PRs carry runner-validated exact matches\n  (plus supporting headers\u002Fdeclarations the matches need to build); they are\n  byte-verified and should read as easy approvals. Fuzzy improvements never go\n  into PRs — they stay on the local branch (the split plan's `local-*` slices)\n  until they become matches. If a match-lane file also carries unshipped fuzzy\n  improvements in other functions, call that out in the PR body.\n- Treat the configured max-files-per-PR as a hard ceiling, not a packing\n  target. Aim for the fewest PRs that are still comfortable to review in one\n  sitting; do not shave slices down to produce a pile of small PRs.\n- Keep broad header, naming, symbol, build, metadata, and rename churn separate\n  from ordinary implementation\u002Fmatching work when possible.\n- Group small changes only when they share one review context and risk class.\n- Treat shared headers, symbol renames, and bot-visible name changes as\n  high-risk even when the file count is small.\n- Avoid path-only splits when a subsystem, actor\u002Fmodule family, matching\n  milestone, API\u002Fheader change, or mechanical cleanup category gives reviewers\n  a clearer unit.\n- Write PR bodies as reviewer-facing digests using the reference template:\n  summary, PR shape, reviewer notes, and verification\u002Fregression status.\n\n## Prepare PR Handoff\n\nUse this path when the user asks to prepare, finalize, refresh, build-check,\nregression-check, update, or hand off a Melee PR.\n\nThe dashboard's `Prepare Handoff` automates the core of this path (pause →\npull upstream & rebase → PR intake → rebuild the production baseline in a\nper-SHA worktree → branch QA vs that baseline → checkpoint with regressed\nsymbols forced to needs_rework → requeue rework at repair priority →\nmatch-only split plan → ship-set verification, where the match-lane diff is\napplied onto the baseline worktree and must build with zero regressions) and\nshares the `\u002Ftmp\u002Fmelee-baseline-\u003Csha>` cache with step 5 below. Use the\nmanual steps for PR-comment triage, PR-body work, and pushing.\n\nFor review-style cleanup and regression triage, use the local QA reference in\n`references\u002Fpr-review-qa-standards.md`. Keep that reference in the workflow\nskill, not in upstream Melee docs.\n\nFor adversarial cleanup of an existing PR, standards-clean source outranks fuzzy\nscore. If removing overzealous worker output, pragmas, fake anchors, dummy\npadding, or other tactic-shaped code lowers fuzzy score or even loses a match,\nkeep the clean source, report the score impact, and route the result through\ncarry-forward or explicit operator policy instead of restoring the tactic.\n\nFor PR splitting\u002Fgrouping and reviewer-facing PR body content, use\n`references\u002Fpr-shaping-reviewer-guidance.md`.\n\n1. Inspect status and PR context:\n\n```bash\ngit status --short --branch --ignore-submodules=all\ngit -C decomp-orchestrator status --short\ngit remote -v\ngh pr status --repo doldecomp\u002Fmelee\n```\n\n2. Sync mainline and missing PR knowledge:\n\n```bash\npython3 decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  --postmortem-scope fetched \\\n  --postmortem-jobs 16\n```\n\n3. Pull active PR comments and reviews:\n\n```bash\ngh pr view --repo doldecomp\u002Fmelee --comments\ngh pr view --repo doldecomp\u002Fmelee --json number,url,title,state,isDraft,baseRefOid,headRefOid,reviewDecision,mergeStateStatus\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fpulls\u002F\u003CPR_NUMBER>\u002Fcomments\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fissues\u002F\u003CPR_NUMBER>\u002Fcomments\ngh api --paginate repos\u002Fdoldecomp\u002Fmelee\u002Fpulls\u002F\u003CPR_NUMBER>\u002Freviews\n```\n\n4. Evaluate PR shape before validation:\n\n- Inventory touched file types and high-risk churn.\n- Check the latest checkpoint (`pr_candidates.md`): only the match candidates\n  ship; notable improvements are listed in `carry_forward.md` and stay local.\n- Decide whether the PR should stay grouped, split, or isolate header\u002Fnaming\n  changes.\n- Draft\u002Fupdate the PR body with a clear summary, PR shape, reviewer notes, and\n  verification section.\n- Call out expected regression false positives caused by renames or declaration\n  movement.\n\n5. Rebuild a baseline from current `origin\u002Fmaster`:\n\n```bash\nBASE_SHA=\"$(git rev-parse origin\u002Fmaster)\"\nBASE_DIR=\"\u002Ftmp\u002Fmelee-baseline-${BASE_SHA}\"\nif [ ! -f \"$BASE_DIR\u002Fbuild\u002FGALE01\u002Fbaseline.json\" ]; then\n  git worktree add --detach \"$BASE_DIR\" \"$BASE_SHA\"\n  (cd \"$BASE_DIR\" && ninja baseline)\nfi\ncp \"$BASE_DIR\u002Fbuild\u002FGALE01\u002Fbaseline.json\" build\u002FGALE01\u002Fbaseline.json\n```\n\n6. Run build and regression gates:\n\n```bash\nninja\nbun run --cwd decomp-orchestrator orch -- \\\n  --repo-root \"$PWD\" \\\n  --state-dir \"$PWD\u002F.decomp-orchestrator-state\u002Fpr-prepare\" \\\n  regression-check \\\n  --target changes_all \\\n  --report-title \"Report for GALE01 PR handoff\" \\\n  --report-max-rows 300\n```\n\n7. Fix and rerun until the handoff state is clean:\n\n- `build\u002FGALE01\u002Fmain.dol: OK`\n- zero broken matches\n- zero unexplained or unaccepted fuzzy regressions in unmatched items\n- zero unit, section, or function metric regressions\n- no unresolved actionable review comments or relevant build warnings\u002Ferrors\n\n8. Run the mandatory pre-ship adversarial review gate (blocks handoff):\n\nSave the match-only split plan as JSON if you do not already have one, then run\nthe pr-review agent in preship mode over every shipping slice:\n\n```bash\nbun run --cwd decomp-orchestrator orch -- \\\n  --project melee \\\n  pr-split-plan --checkpoint \u003Ccheckpoint.json> --ship-status \u003Cship_status.json> --json \\\n  > \u002Ftmp\u002Fmelee-pr-split-plan.json\nbun run --cwd decomp-orchestrator orch -- \\\n  --project melee \\\n  pr-preship-review --plan \u002Ftmp\u002Fmelee-pr-split-plan.json --all\n```\n\n- ANY `reject` finding blocks handoff; exit 1 means do not draft PR bodies or\n  push. Agent\u002Ftool failures also exit 1 (the gate fails closed) — fix the\n  infrastructure and rerun rather than skipping the gate.\n- Disposition of rejects, consistent with the MATCHES-only promotion policy:\n  mark the affected symbols `needs_rework`, requeue them at repair priority,\n  and ship the slice without them or not at all. Never keep a rejected hunk to\n  preserve a match score.\n- Per-slice verdicts, findings, and prompts land under\n  `decomp-orchestrator\u002Fprojects\u002Fmelee\u002Fstate\u002Fpreship_reviews\u002F\u003Crun-id>\u002F\u003Cslice-id>\u002F`\n  (`review.md` is the human-readable digest); cite them in the PR body's\n  verification section.\n\n9. Commit or amend only parent Melee PR files, then push safely:\n\n```bash\ngit status --short --untracked-files=no --ignore-submodules=all\ngit diff --check\ngit add \u003Cmelee-source-files>\ngit commit --amend --no-edit\ngit push --force-with-lease fork HEAD:\u003Cbranch>\n```\n\n10. Update or create the PR, then watch CI:\n\n```bash\ngh pr edit \u003CPR_NUMBER> --repo doldecomp\u002Fmelee --body-file \u003Creport.md>\ngh pr checks \u003CPR_NUMBER> --repo doldecomp\u002Fmelee\ngh run view \u003CRUN_ID> --repo doldecomp\u002Fmelee --log-failed\n```\n\nInclude PR URL, head SHA, base SHA, changed units, match\u002Fregression summary,\nlocal gate results, and CI state in the final response and PR body.\n\n## Verification\n\nFor skill maintenance or command sanity checks:\n\n```bash\npython3 -m py_compile \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fsync_repo_and_pr_library.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Ffetch_recent_pr_dump.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fbuild_pr_postmortems.py \\\n  decomp-orchestrator\u002Fknowledge\u002Fsources\u002Fcode_context\u002Fpast_prs\u002Fcommands\u002Fmigrate_pr_data_layout.py\n```\n",{"name":5,"description":7},"main","e7e6fb660782e780949acacded7546400b54f2ae",".codex\u002Fskills\u002Fmelee-pr-workflow",3,"https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness","https:\u002F\u002Fgithub.com\u002Ffjooord","https:\u002F\u002Flascari.ai\u002F","2026-11-06 04:39:50",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add fjooord\u002Fgamecube-decomp-harness --skill melee-pr-workflow",{"tool":38,"command":39},"bunx","bunx skills add fjooord\u002Fgamecube-decomp-harness --skill melee-pr-workflow",{"tool":41,"command":42},"pnpm","pnpm dlx skills add fjooord\u002Fgamecube-decomp-harness --skill melee-pr-workflow","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fmelee-pr-workflow\u002FSKILL.md",true,[46,52,56],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"agents\u002Fopenai.yaml","REFERENCE","参考文档",325,"https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fmelee-pr-workflow\u002Fagents\u002Fopenai.yaml",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fpr-review-qa-standards.md",3447,"https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fmelee-pr-workflow\u002Freferences\u002Fpr-review-qa-standards.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fpr-shaping-reviewer-guidance.md",5375,"https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fmelee-pr-workflow\u002Freferences\u002Fpr-shaping-reviewer-guidance.md",[61,68,75,82,89,96],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M78ST6YK3EBH0GKFER68ZV8V","better-writing","UX writing and interface copy, from voice and button labels to error messages and empty states. Use when writing or reviewing any user-facing text: button and link labels, form errors, placeholders, settings labels, onboarding flows, notifications, or empty states. Triggers on UX writing, microcopy, interface copy, product copy, copywriting, button labels, link text, error messages, empty states, placeholder text, settings labels, capitalization, title case, sentence case, voice and tone.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-writing\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M78ST15FSPN2WB4T98BP4HXE","better-ui","Design engineering principles for making interfaces feel polished. Use when building UI components, reviewing frontend code, implementing animations, hover states, shadows, borders, micro-interactions, enter\u002Fexit animations, choosing or reviewing icons, or any visual detail work. Triggers on UI polish, design details, \"make it feel better\", \"feels off\", stagger animations, border radius, optical alignment, image outlines, box shadows, icons, icon stroke weight, icon states, motion restraint.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-ui\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M78SSSVHD4MB3GWJFPHAR6NX","better-typography","Web typography from choosing fonts to spacing, wrapping and accessibility. Use when picking or pairing typefaces, configuring variable fonts or OpenType features, setting up a type scale, checking heading hierarchy, styling text in components, truncating text, styling underlines, selection, placeholders or carets, or reviewing frontend code for typography. Triggers on typography, fonts, font formats, woff2, variable fonts, font-weight, opentype, font-feature-settings, letter-spacing, line-height, type scale, heading hierarchy, heading levels, tabular numbers, text-wrap, truncation, line clamp, underlines, text-decoration, text selection, iOS input zoom, font smoothing, text contrast, measure, line length, text-box, smart punctuation, drop cap.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-typography\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M78SSM8Z4ESHX0YA5GQSKREK","better-layout","Layout structure for web interfaces, from grouping and alignment to reading order, progressive disclosure, and adaptive breakpoints. Use when structuring a page or component, spacing or aligning controls, deciding what collapses at small sizes, handling RTL layout direction, or reviewing frontend code for layout. Triggers on layout, spacing, alignment, grouping, negative space, whitespace, visual hierarchy, reading order, progressive disclosure, breakpoints, responsive layout, container queries, safe area, full-bleed, edge-to-edge, layout margins, RTL layout, logical properties.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-layout\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M78SSFS2YJ7CAJTS6PXA8YYY","better-interface","User-invoked, cross-discipline interface review that coordinates better-accessibility, better-layout, better-writing, better-typography, better-colors, and better-ui. Use when explicitly invoked for a holistic review of a screen, flow, feature, or product interface. Supports quick and full review modes. Triggers on better-interface, full interface review, holistic UI audit, cross-discipline design review, review the whole interface.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-interface\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":19},"01M78SSB3HCMJVWVZYQ7QN5VPX","better-colors","OKLCH color space and color usage for web projects. Convert hex\u002Frgb\u002Fhsl to oklch, generate palettes, check contrast, handle gamut boundaries, theme with Tailwind v4, and apply color with meaning. Triggers on oklch, color conversion, palette generation, contrast ratio, gamut, display p3, design tokens, semantic color tokens, hue drift, chroma, dark mode colors, accent color, color meaning, light and dark appearance, increased contrast.","https:\u002F\u002Fgithub.com\u002Ffjooord\u002Fgamecube-decomp-harness\u002Fblob\u002Fmain\u002Fref\u002Fjakubkrehel skills main skills\u002Fbetter-colors\u002FSKILL.md",[],[]]