[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75VT42QPTFSHMXHCEQVVJYN":3,"skill-same:01M75VT42QPTFSHMXHCEQVVJYN":47},{"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":31,"updatedAt":32,"install":33,"resources":46},"01M75VT42QPTFSHMXHCEQVVJYN","pr-fix",null,"Bring a PR to mergeable — fix CI failures, address review comments, and resolve merge conflicts. Runs either from a supplied watcher worklist (delegated mode, one pass) or from its own assessment (direct mode, looping until clear). Use after opening a PR when CI is red, reviewers have left comments, or the branch conflicts with its base.","ronaknnathani\u002Frelay","ronaknnathani","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F7279934?v=4","01M75VS2S5V1DVP31PCWYA7SMA","Ronak Nathani","01M75VS2STNTX5Y5JGYQGRE0NG",44,"MIT","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fpr-fix\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: pr-fix\ndescription: Bring a PR to mergeable — fix CI failures, address review comments, and resolve merge conflicts. Runs either from a supplied watcher worklist (delegated mode, one pass) or from its own assessment (direct mode, looping until clear). Use after opening a PR when CI is red, reviewers have left comments, or the branch conflicts with its base.\n---\n\n# PR Fix\n\nDrive a PR to a mergeable state: a green CI, every review comment fixed-and-resolved or\nreplied-and-flagged, and no merge conflicts with the base. Work the three\nfronts — CI, comments, conflicts — and loop until all are clear. Fix root causes, never silence\nfailures or guess an author's intent. Use `review`'s shared severity vocabulary (Critical \u002F Important \u002F\nSuggestion). Run independent investigations as sub-agents when available; otherwise do them inline.\n\n## Two modes — check your input first\n\n**Delegated mode — a caller supplied a watcher worklist.** `pr-monitor` hands you items taken from a\n`relay pr watch` digest: each carries `reason`, `source`, `id`, `answers`, `updatedAt`, `body`,\n`thread_id`, `path`, `line`, and for checks the `check_name` and `check_run_id`. That worklist is **complete and\nauthoritative**:\n\n- **Skip step 1's broad assessment.** Do not re-run the PR\u002Fcomment\u002Fthread\u002Fcheck sweep and do not build\n  a full context bundle — the watcher already observed all of it. Bind `REPO` and `PR` and go straight\n  to the work.\n- Fetch only what the specific item needs: the failed run's log for a `check_run_id`, `git log`\u002F`git\n  blame` for a conflicting hunk, the file under an inline comment.\n- **Fix once, then return** — no reassessment loop. The watcher re-observes after your push, and the\n  next attention event carries whatever is left.\n- **Return a structured result, one entry per supplied item:** the item id, what you did, `fixed`,\n  `replied`, `escalated`, `ignored_non_actionable`, or `failed`, and the reason when it is not\n  `fixed`. Report pushed commits and the new head SHA. `ignored_non_actionable` means no reply,\n  reaction, resolution, or other GitHub mutation was made for that item.\n- **Never** run `relay pr watch tick` or `status`, and never schedule anything. The caller owns the\n  watcher record.\n\n**Direct mode — no worklist was supplied.** Assess the PR yourself and loop until clear, exactly as\ndescribed below. This is the manual path and it is unchanged.\n\n## Quick commands\n\n| Task | Command |\n|------|---------|\n| PR + check status | `gh pr view --json number,title,state,statusCheckRollup,mergeable` |\n| CI checks | `gh pr checks` |\n| Failed run logs | `gh run view \u003CRUN_ID> --log-failed` |\n| Inline review comments | `gh api \"repos\u002F$REPO\u002Fpulls\u002F$PR\u002Fcomments\" --jq '.[] \\| {id,user:.user.login,path,line,body}'` |\n| Reply to an inline comment | `gh api \"repos\u002F$REPO\u002Fpulls\u002Fcomments\u002F\u003CID>\u002Freplies\" -f body=\"$BODY\"` |\n| Reply to a conversation comment | `gh pr comment \"$PR\" --body \"$BODY\"` |\n| Reply to a review body | `gh pr review \"$PR\" --comment --body \"$BODY\"` |\n| Resolve a thread (after fixing) | `gh api graphql -f query='mutation($t:ID!){resolveReviewThread(input:{threadId:$t}){thread{isResolved}}}' -F t=\u003CTHREAD_ID>` |\n\n`\u003CRUN_ID>` comes from a delegated item's `check_run_id`, or from `gh pr checks` \u002F `statusCheckRollup`;\na `\u003CTHREAD_ID>` comes from a delegated item's `thread_id`, or from the GraphQL `reviewThreads` query.\n`gh` has no native thread-resolve — resolving requires the GraphQL mutation above.\n\n## Every automated reply — exact marker, visible disclosure, same source\n\nEvery reply you post to a pull request starts with these two lines, then a blank line, then the reply:\n\n```\n\u003C!-- relay-agent-reply answers=\u003Citem answers token> -->\n🤖 \u003Cagent> on behalf of \u003Cauthor>\n```\n\nThe HTML comment is invisible on GitHub and is the **only** thing that tells the `relay pr watch`\nruntime an agent wrote a reply. A reply without it is read as fresh human feedback and will wake the\nowner again forever; the emoji line alone is not enough, because a human can type an emoji. The\ndisclosure line is the human-visible half of the same promise: never write as if you were the author.\n\n**The marker must name the exact item it answers — copy the item's `answers` field verbatim.** Every\ndigest item carries one, and the watcher matches replies by that id, never by time. A marker that\nnames nothing (`\u003C!-- relay-agent-reply answers= -->` or the bare marker) answers nothing on a\nconversation, a review, or a thread, so the item keeps waking the owner. A marker that names the\nwrong id answers the wrong thing — and never the one in front of you.\n\nThis is not bookkeeping. A reviewer can write a second comment between the watcher's last look and\nyour reply, and the watcher has never shown it to anybody: a reply anchored to the id you were given\nleaves that comment actionable, while an unanchored one buries it.\n\n**Reply on the same source you are answering.** The watcher reconciles each source independently, so\nan answer posted somewhere else does not answer anything:\n\n| Item | Reply with | Marker |\n|---|---|---|\n| `new-comment` | `gh pr comment` — the pull request conversation | `answers=comment:200` |\n| `new-review` | `gh pr review --comment` — a review, not a conversation comment | `answers=review:100` |\n| `new-inline-comment` | the inline replies endpoint on that comment | `answers=inline-comment:300` |\n| `unresolved-thread` | the inline replies endpoint on that thread | `answers=review-thread:\u003Cthread-id>:\u003Ccomment-id>` |\n| `changes-requested` | `gh pr review --comment` on the review that requested them | `answers=review:100` |\n\nThe ids in that column are examples; the item's own `answers` token is the truth. A thread's token\nnames the exact comment the digest reported, because a new reply arriving beside it is a different\nitem the watcher must still be able to surface.\n\n## Process\n\n1. **Assess.** *(Direct mode only — delegated mode skips this entirely and uses the supplied\n   worklist.)* `REPO=$(gh repo view --json nameWithOwner --jq .nameWithOwner)`,\n   `PR=$(gh pr view --json number --jq .number)`. Pull check status, the comment list, and `mergeable`.\n   Triage into the three fronts below. Detect the repo's OWN build\u002Ftest\u002Flint commands from its\n   `Makefile`, `package.json` scripts, or CI config (`.github\u002Fworkflows\u002F*`) — never assume a toolchain.\n\n   Before editing, materialize the remote PR context onto the local filesystem so the fix is based on a\n   stable, inspectable record rather than scattered terminal output. Use a non-committed directory under\n   the git metadata dir, e.g. `CTX_DIR=\"$(git rev-parse --git-dir)\u002Frelay\u002Fpr-fix\u002F$PR\"`, and write at\n   least: PR metadata, `statusCheckRollup`, `gh pr checks`, inline review comments, review threads,\n   failed-run logs, the base-vs-head diff, and the discovered build\u002Ftest commands. Re-read those files to\n   plan the fix. Refresh the bundle after every push before deciding the PR is clear.\n\n2. **CI failures — Stop-the-Line, red-loop-first.** Do not blindly do what the error message literally\n   says; diagnose. For each failing check:\n   - **Reproduce locally first.** Run the repo's own failing command to get a tight red loop. If it\n     only fails in CI, line up versions\u002Fenv before guessing.\n   - **Localize, then minimize.** Find the failing layer (triage table below), then shrink to the\n     smallest failing case — one test, one file.\n   - **Fix the ROOT cause**, not the symptom.\n   - **Regression test: red before, green after.** Add\u002Fadjust a test that fails without your fix and\n     passes with it. Re-run the repo's full check command end to end until green.\n   - **Never silence a failure** — no deleting or skipping a test, no lint-suppression, no loosened\n     assertion to go green. A test that now fails means behavior changed: fix the code, not the test.\n\n   | Category | Tell | Route |\n   |---|---|---|\n   | Type | type\u002Fcompile error, signature mismatch | fix the type or the call site at the root |\n   | Import | unresolved\u002Fcircular import, missing symbol | fix the path\u002Fexport; do not stub it out |\n   | Config | failing lint\u002Fformat\u002FCI step config | match the repo's configured rule, don't suppress |\n   | Dependency | version\u002Flockfile\u002Fresolution error | reconcile the manifest + lockfile together |\n   | Environment | passes locally, fails only in CI | align runtime\u002Fversion\u002Fenv vars with CI |\n\n3. **Review comments — classify each, then act.**\n   - **Non-actionable** (praise, thanks, FYI\u002Fstatus-only notes, approvals, duplicate summaries,\n     automated \"no findings\" \u002F \"no important findings\" reports, or any comment with no concrete\n     defect, question, requested change, requested decision, or blocker): do nothing on GitHub. Do\n     not post an acknowledgment such as \"Acknowledged\", do not react, and do not resolve an\n     unresolved thread merely to clear it. In delegated mode return `ignored_non_actionable` with a\n     short reason; in direct mode exclude it from the actionable worklist.\n   - **Obvious gap-fix** (a clear bug, a missing test, a style\u002Frule violation, a one-right-answer\n     mechanical change): implement it, reply describing exactly what changed, and resolve the thread\n     (GraphQL `resolveReviewThread` — see Quick commands). Only mark a thread resolved once the fix is\n     pushed; if you can't resolve it programmatically, say so rather than claiming it's done.\n   - **Author decision** (changes intended behavior, API shape, or scope; two-plus reasonable answers):\n     do NOT guess. Reply asking for input, FLAG it to the author, and leave the thread open.\n   - **When unsure which it is, treat it as a decision** and surface it.\n   - Every reply carries the marker naming that item's `answers` token and the visible\n     `🤖 \u003Cagent> on behalf of \u003Cauthor>` disclosure, and goes on the same source it answers.\n\n4. **Merge conflicts — research both intents, never abort.** Rebase\u002Fmerge onto the base. For each\n   conflict, research the intent behind BOTH sides before resolving — read the commit messages and the\n   PR that introduced each hunk (`git log`, `git blame`, `gh pr view`). Preserve both intents where\n   feasible. NEVER `--abort` the rebase\u002Fmerge; resolve forward. Re-run the repo's validation after\n   resolving and before continuing. Then **confirm your work survived** — your branch's commits are\n   still in `git log` and the net diff against the base still contains your intended changes; a\n   resolve-forward can silently drop a hunk even when validation passes.\n\n5. **Loop until clear.** *(Direct mode only.)* Commit and push fixes, then re-assess (step 1). Repeat\n   until CI is green and every actionable comment is addressed — fixed+resolved, or\n   replied+flagged. Non-actionable comments require no acknowledgment and do not block completion.\n   Surface the flagged decisions to the caller as the remaining blockers. **In delegated mode, stop\n   after one pass** and return the per-item result instead; the watcher re-observes and the caller\n   decides what happens next.\n\n## Red flags\n\n- Doing what the error message literally says without diagnosing the root cause.\n- Deleting\u002Fskipping a test, suppressing a lint, or weakening an assertion to turn a check green.\n- Editing a failing test instead of the code it caught — a red test means behavior changed.\n- Guessing an author-decision comment instead of replying and flagging it.\n- Replying to a non-actionable comment just to acknowledge it, including \"no findings\" or \"no\n  important findings\" review summaries.\n- A reply that reads as the human author's own words, with no automated-agent disclosure.\n- A reply missing the marker, naming no `answers` token, naming an id you invented instead of the\n  item's own, or posted on a different source than the one it answers — every one of those either\n  keeps waking the owner or answers something nobody asked about.\n- Running `git rebase --abort` \u002F `git merge --abort` instead of resolving the conflict.\n- Resolving a conflict by keeping one side without understanding why the other side exists.\n- Assuming `npm`\u002F`make`\u002Fetc. instead of the command the repo's own config actually uses.\n- Fixing from transient terminal output instead of a local PR context bundle that can be re-read and\n  refreshed. *(Direct mode; in delegated mode the supplied worklist is that record.)*\n- Re-fetching the whole PR context, or looping, when a delegated worklist was supplied.\n- Touching the watcher record — `relay pr watch tick` and `status` belong to the caller.\n\n## Verification checklist\n\n- [ ] In delegated mode: every supplied item has a returned outcome, no broad re-fetch or reassessment loop ran, and the watcher record was left untouched.\n- [ ] In direct mode: `gh pr checks` is fully green; each fix reproduced a local red loop and has a red-before\u002Fgreen-after regression test.\n- [ ] Direct mode captured the remote PR context locally before edits and refreshed it after each push.\n- [ ] No failure was silenced (no skipped\u002Fdeleted test, no lint-suppression, no loosened assertion).\n- [ ] Every actionable review comment is fixed+resolved, or replied+flagged as an author decision\n      left open; non-actionable comments received no reply, reaction, or resolution.\n- [ ] Every agent reply carries the marker with that item's exact `answers` token and the visible disclosure, and was posted on the same source it answers.\n- [ ] All conflicts resolved forward (no abort), both intents researched and preserved, validation re-run after.\n- [ ] After any rebase, confirmed my branch's commits survived (still in `git log`; net diff vs base still carries my changes).\n- [ ] Findings reported with the shared severity vocabulary (Critical \u002F Important \u002F Suggestion).\n","# PR Fix\n\nDrive a PR to a mergeable state: a green CI, every review comment fixed-and-resolved or\nreplied-and-flagged, and no merge conflicts with the base. Work the three\nfronts — CI, comments, conflicts — and loop until all are clear. Fix root causes, never silence\nfailures or guess an author's intent. Use `review`'s shared severity vocabulary (Critical \u002F Important \u002F\nSuggestion). Run independent investigations as sub-agents when available; otherwise do them inline.\n\n## Two modes — check your input first\n\n**Delegated mode — a caller supplied a watcher worklist.** `pr-monitor` hands you items taken from a\n`relay pr watch` digest: each carries `reason`, `source`, `id`, `answers`, `updatedAt`, `body`,\n`thread_id`, `path`, `line`, and for checks the `check_name` and `check_run_id`. That worklist is **complete and\nauthoritative**:\n\n- **Skip step 1's broad assessment.** Do not re-run the PR\u002Fcomment\u002Fthread\u002Fcheck sweep and do not build\n  a full context bundle — the watcher already observed all of it. Bind `REPO` and `PR` and go straight\n  to the work.\n- Fetch only what the specific item needs: the failed run's log for a `check_run_id`, `git log`\u002F`git\n  blame` for a conflicting hunk, the file under an inline comment.\n- **Fix once, then return** — no reassessment loop. The watcher re-observes after your push, and the\n  next attention event carries whatever is left.\n- **Return a structured result, one entry per supplied item:** the item id, what you did, `fixed`,\n  `replied`, `escalated`, `ignored_non_actionable`, or `failed`, and the reason when it is not\n  `fixed`. Report pushed commits and the new head SHA. `ignored_non_actionable` means no reply,\n  reaction, resolution, or other GitHub mutation was made for that item.\n- **Never** run `relay pr watch tick` or `status`, and never schedule anything. The caller owns the\n  watcher record.\n\n**Direct mode — no worklist was supplied.** Assess the PR yourself and loop until clear, exactly as\ndescribed below. This is the manual path and it is unchanged.\n\n## Quick commands\n\n| Task | Command |\n|------|---------|\n| PR + check status | `gh pr view --json number,title,state,statusCheckRollup,mergeable` |\n| CI checks | `gh pr checks` |\n| Failed run logs | `gh run view \u003CRUN_ID> --log-failed` |\n| Inline review comments | `gh api \"repos\u002F$REPO\u002Fpulls\u002F$PR\u002Fcomments\" --jq '.[] \\| {id,user:.user.login,path,line,body}'` |\n| Reply to an inline comment | `gh api \"repos\u002F$REPO\u002Fpulls\u002Fcomments\u002F\u003CID>\u002Freplies\" -f body=\"$BODY\"` |\n| Reply to a conversation comment | `gh pr comment \"$PR\" --body \"$BODY\"` |\n| Reply to a review body | `gh pr review \"$PR\" --comment --body \"$BODY\"` |\n| Resolve a thread (after fixing) | `gh api graphql -f query='mutation($t:ID!){resolveReviewThread(input:{threadId:$t}){thread{isResolved}}}' -F t=\u003CTHREAD_ID>` |\n\n`\u003CRUN_ID>` comes from a delegated item's `check_run_id`, or from `gh pr checks` \u002F `statusCheckRollup`;\na `\u003CTHREAD_ID>` comes from a delegated item's `thread_id`, or from the GraphQL `reviewThreads` query.\n`gh` has no native thread-resolve — resolving requires the GraphQL mutation above.\n\n## Every automated reply — exact marker, visible disclosure, same source\n\nEvery reply you post to a pull request starts with these two lines, then a blank line, then the reply:\n\n```\n\u003C!-- relay-agent-reply answers=\u003Citem answers token> -->\n🤖 \u003Cagent> on behalf of \u003Cauthor>\n```\n\nThe HTML comment is invisible on GitHub and is the **only** thing that tells the `relay pr watch`\nruntime an agent wrote a reply. A reply without it is read as fresh human feedback and will wake the\nowner again forever; the emoji line alone is not enough, because a human can type an emoji. The\ndisclosure line is the human-visible half of the same promise: never write as if you were the author.\n\n**The marker must name the exact item it answers — copy the item's `answers` field verbatim.** Every\ndigest item carries one, and the watcher matches replies by that id, never by time. A marker that\nnames nothing (`\u003C!-- relay-agent-reply answers= -->` or the bare marker) answers nothing on a\nconversation, a review, or a thread, so the item keeps waking the owner. A marker that names the\nwrong id answers the wrong thing — and never the one in front of you.\n\nThis is not bookkeeping. A reviewer can write a second comment between the watcher's last look and\nyour reply, and the watcher has never shown it to anybody: a reply anchored to the id you were given\nleaves that comment actionable, while an unanchored one buries it.\n\n**Reply on the same source you are answering.** The watcher reconciles each source independently, so\nan answer posted somewhere else does not answer anything:\n\n| Item | Reply with | Marker |\n|---|---|---|\n| `new-comment` | `gh pr comment` — the pull request conversation | `answers=comment:200` |\n| `new-review` | `gh pr review --comment` — a review, not a conversation comment | `answers=review:100` |\n| `new-inline-comment` | the inline replies endpoint on that comment | `answers=inline-comment:300` |\n| `unresolved-thread` | the inline replies endpoint on that thread | `answers=review-thread:\u003Cthread-id>:\u003Ccomment-id>` |\n| `changes-requested` | `gh pr review --comment` on the review that requested them | `answers=review:100` |\n\nThe ids in that column are examples; the item's own `answers` token is the truth. A thread's token\nnames the exact comment the digest reported, because a new reply arriving beside it is a different\nitem the watcher must still be able to surface.\n\n## Process\n\n1. **Assess.** *(Direct mode only — delegated mode skips this entirely and uses the supplied\n   worklist.)* `REPO=$(gh repo view --json nameWithOwner --jq .nameWithOwner)`,\n   `PR=$(gh pr view --json number --jq .number)`. Pull check status, the comment list, and `mergeable`.\n   Triage into the three fronts below. Detect the repo's OWN build\u002Ftest\u002Flint commands from its\n   `Makefile`, `package.json` scripts, or CI config (`.github\u002Fworkflows\u002F*`) — never assume a toolchain.\n\n   Before editing, materialize the remote PR context onto the local filesystem so the fix is based on a\n   stable, inspectable record rather than scattered terminal output. Use a non-committed directory under\n   the git metadata dir, e.g. `CTX_DIR=\"$(git rev-parse --git-dir)\u002Frelay\u002Fpr-fix\u002F$PR\"`, and write at\n   least: PR metadata, `statusCheckRollup`, `gh pr checks`, inline review comments, review threads,\n   failed-run logs, the base-vs-head diff, and the discovered build\u002Ftest commands. Re-read those files to\n   plan the fix. Refresh the bundle after every push before deciding the PR is clear.\n\n2. **CI failures — Stop-the-Line, red-loop-first.** Do not blindly do what the error message literally\n   says; diagnose. For each failing check:\n   - **Reproduce locally first.** Run the repo's own failing command to get a tight red loop. If it\n     only fails in CI, line up versions\u002Fenv before guessing.\n   - **Localize, then minimize.** Find the failing layer (triage table below), then shrink to the\n     smallest failing case — one test, one file.\n   - **Fix the ROOT cause**, not the symptom.\n   - **Regression test: red before, green after.** Add\u002Fadjust a test that fails without your fix and\n     passes with it. Re-run the repo's full check command end to end until green.\n   - **Never silence a failure** — no deleting or skipping a test, no lint-suppression, no loosened\n     assertion to go green. A test that now fails means behavior changed: fix the code, not the test.\n\n   | Category | Tell | Route |\n   |---|---|---|\n   | Type | type\u002Fcompile error, signature mismatch | fix the type or the call site at the root |\n   | Import | unresolved\u002Fcircular import, missing symbol | fix the path\u002Fexport; do not stub it out |\n   | Config | failing lint\u002Fformat\u002FCI step config | match the repo's configured rule, don't suppress |\n   | Dependency | version\u002Flockfile\u002Fresolution error | reconcile the manifest + lockfile together |\n   | Environment | passes locally, fails only in CI | align runtime\u002Fversion\u002Fenv vars with CI |\n\n3. **Review comments — classify each, then act.**\n   - **Non-actionable** (praise, thanks, FYI\u002Fstatus-only notes, approvals, duplicate summaries,\n     automated \"no findings\" \u002F \"no important findings\" reports, or any comment with no concrete\n     defect, question, requested change, requested decision, or blocker): do nothing on GitHub. Do\n     not post an acknowledgment such as \"Acknowledged\", do not react, and do not resolve an\n     unresolved thread merely to clear it. In delegated mode return `ignored_non_actionable` with a\n     short reason; in direct mode exclude it from the actionable worklist.\n   - **Obvious gap-fix** (a clear bug, a missing test, a style\u002Frule violation, a one-right-answer\n     mechanical change): implement it, reply describing exactly what changed, and resolve the thread\n     (GraphQL `resolveReviewThread` — see Quick commands). Only mark a thread resolved once the fix is\n     pushed; if you can't resolve it programmatically, say so rather than claiming it's done.\n   - **Author decision** (changes intended behavior, API shape, or scope; two-plus reasonable answers):\n     do NOT guess. Reply asking for input, FLAG it to the author, and leave the thread open.\n   - **When unsure which it is, treat it as a decision** and surface it.\n   - Every reply carries the marker naming that item's `answers` token and the visible\n     `🤖 \u003Cagent> on behalf of \u003Cauthor>` disclosure, and goes on the same source it answers.\n\n4. **Merge conflicts — research both intents, never abort.** Rebase\u002Fmerge onto the base. For each\n   conflict, research the intent behind BOTH sides before resolving — read the commit messages and the\n   PR that introduced each hunk (`git log`, `git blame`, `gh pr view`). Preserve both intents where\n   feasible. NEVER `--abort` the rebase\u002Fmerge; resolve forward. Re-run the repo's validation after\n   resolving and before continuing. Then **confirm your work survived** — your branch's commits are\n   still in `git log` and the net diff against the base still contains your intended changes; a\n   resolve-forward can silently drop a hunk even when validation passes.\n\n5. **Loop until clear.** *(Direct mode only.)* Commit and push fixes, then re-assess (step 1). Repeat\n   until CI is green and every actionable comment is addressed — fixed+resolved, or\n   replied+flagged. Non-actionable comments require no acknowledgment and do not block completion.\n   Surface the flagged decisions to the caller as the remaining blockers. **In delegated mode, stop\n   after one pass** and return the per-item result instead; the watcher re-observes and the caller\n   decides what happens next.\n\n## Red flags\n\n- Doing what the error message literally says without diagnosing the root cause.\n- Deleting\u002Fskipping a test, suppressing a lint, or weakening an assertion to turn a check green.\n- Editing a failing test instead of the code it caught — a red test means behavior changed.\n- Guessing an author-decision comment instead of replying and flagging it.\n- Replying to a non-actionable comment just to acknowledge it, including \"no findings\" or \"no\n  important findings\" review summaries.\n- A reply that reads as the human author's own words, with no automated-agent disclosure.\n- A reply missing the marker, naming no `answers` token, naming an id you invented instead of the\n  item's own, or posted on a different source than the one it answers — every one of those either\n  keeps waking the owner or answers something nobody asked about.\n- Running `git rebase --abort` \u002F `git merge --abort` instead of resolving the conflict.\n- Resolving a conflict by keeping one side without understanding why the other side exists.\n- Assuming `npm`\u002F`make`\u002Fetc. instead of the command the repo's own config actually uses.\n- Fixing from transient terminal output instead of a local PR context bundle that can be re-read and\n  refreshed. *(Direct mode; in delegated mode the supplied worklist is that record.)*\n- Re-fetching the whole PR context, or looping, when a delegated worklist was supplied.\n- Touching the watcher record — `relay pr watch tick` and `status` belong to the caller.\n\n## Verification checklist\n\n- [ ] In delegated mode: every supplied item has a returned outcome, no broad re-fetch or reassessment loop ran, and the watcher record was left untouched.\n- [ ] In direct mode: `gh pr checks` is fully green; each fix reproduced a local red loop and has a red-before\u002Fgreen-after regression test.\n- [ ] Direct mode captured the remote PR context locally before edits and refreshed it after each push.\n- [ ] No failure was silenced (no skipped\u002Fdeleted test, no lint-suppression, no loosened assertion).\n- [ ] Every actionable review comment is fixed+resolved, or replied+flagged as an author decision\n      left open; non-actionable comments received no reply, reaction, or resolution.\n- [ ] Every agent reply carries the marker with that item's exact `answers` token and the visible disclosure, and was posted on the same source it answers.\n- [ ] All conflicts resolved forward (no abort), both intents researched and preserved, validation re-run after.\n- [ ] After any rebase, confirmed my branch's commits survived (still in `git log`; net diff vs base still carries my changes).\n- [ ] Findings reported with the shared severity vocabulary (Critical \u002F Important \u002F Suggestion).\n",{"name":5,"description":7},"main","5ab51c2d8af79f04294f91621a5b65e21f39dd53","skills\u002Fpr-fix",4,"https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay","https:\u002F\u002Fgithub.com\u002Fronaknnathani","ronaknathani.com","Compute Infrastructure | Kubernetes @ LinkedIn","2026-11-06 14:25:14",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add ronaknnathani\u002Frelay --skill pr-fix",{"tool":39,"command":40},"bunx","bunx skills add ronaknnathani\u002Frelay --skill pr-fix",{"tool":42,"command":43},"pnpm","pnpm dlx skills add ronaknnathani\u002Frelay --skill pr-fix","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fpr-fix\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M75VVBP3R7RHF93R3RFWC3SE","write-like-me","Write or edit prose in author's voice, or check that a draft sounds like them. Use for blog posts, design and strategy docs, PR descriptions, commits and review comments, chat \u002F agent instructions, peer feedback, or any first-person writing on their behalf, and whenever an agent is producing text that will go out under their name and needs to sound like them, not like a generated reply.","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fwrite-like-me\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M75VV70MDT9DZWQG1ZCA3VH6","validate","Verify a change is ready to ship by checking it against its acceptance criteria and running the repo's own quality gates in order, then report a pass\u002Ffail verdict with the exact failures. Use to decide go\u002Fno-go right before shipping — when you want a pass\u002Ffail verdict against acceptance criteria and the repo's quality gates (lint\u002Fbuild\u002Ftest), not a code critique (use `review` for that). It reports — it does not fix (route failures to `pr-fix`\u002F`implement`) and does not ship.","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fvalidate\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M75VV5DEBH4ZS4XDY27JGHCF","todo","Capture, list, and complete repo-scoped todos","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Ftodo\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M75VV38NYR4V51C27WPMFG5F","tl","Manage a Relay program as the CEO-facing tech lead: shape goals and contracts, prioritize dependency-aware work, dispatch senior worker agents, and surface every decision without writing production code or merging.","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Ftl\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M75VTY2ZTZYJ09AYP6N46FRT","stack-ship","Use only for a standalone stack explicitly launched with `relay --workflow stack-ship`; never invoke inside a tech-lead-managed Relay program. Decompose one goal into small PRs, delegate build and monitor work, surface author decisions, and auto-merge only after human approval.","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fstack-ship\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M75VTS6YN6082J93ASWJKENW","simplify","Reduce complexity in recently changed code — cut needless abstraction, dead code, and duplication; improve names and structure — WITHOUT changing behavior. Quality only; it does not hunt for bugs (use review for that). Use after implementing a chunk of work, before review, or whenever a diff feels more complicated than the problem it solves. Triggers on \"clean this up\" \u002F \"tidy this\" \u002F \"refactor this\" when code already exists; for an underspecified goal with no code yet, use `clarify`.","https:\u002F\u002Fgithub.com\u002Fronaknnathani\u002Frelay\u002Fblob\u002Fmain\u002Fskills\u002Fsimplify\u002FSKILL.md",[],[]]