github-pr-shepherding
Skill0 starsChanged 3 days agoMITView on GitHub
Use when monitoring, unblocking, or driving GitHub PRs or native stacks through required checks, group merge, merge queues, auto-merge for ordinary PRs, runner pickup, or blocker comments. Verify current GitHub state and report merged or blocked with evidence.
- Reads credentials
In this repository108 other files it gives its agents.
- adaptive-estimationSkill
- adaptive-loop-budgetingSkill
- adopt-and-deleteSkill
- adversarial-reviewSkill
- agent-drives-uiSkill
- agentic-coding-loopSkill
- agent-otel-trajectorySkill
- agents-as-iac-modulesSkill
- ascii-ui-mockupsSkill
- authoring-html-emailSkill
- authoring-skills-vs-mcpSkill
- batched-ssh-enumerationSkill
- c4-architectureSkill
- challenge-the-premiseSkill
- codegraph-bootstrapSkill
- codegraph-worktree-startupSkill
- colocated-pod-tolerationsSkill
- command-creatorSkill
- complexity-budgetsSkill
- connector-readinessSkill
- context-anchored-patchingSkill
- continuous-issue-resolutionSkill
- credential-proxySkill
- criticality-tiered-reportingSkill
- data-connector-buildingSkill
- declarative-firstSkill
- disk-pressure-forensicsSkill
- early-return-over-elseSkill
- email-triageSkill
- enums-codify-behaviorSkill
- ephemeral-workspace-lifecycleSkill
- event-sourcingSkill
- evidence-claim-boundariesSkill
- exhaustive-matchSkill
- feature-coverage-not-just-line-coverageSkill
- full-disk-forensicsSkill
- gate-before-pushSkill
- graceful-shutdown-stateful-agentsSkill
- grounded-generationSkill
- iac-incident-responseSkill
- iac-not-ad-hocSkill
- instrumented-service-scaffoldSkill
- interface-over-conditionalsSkill
- knowledge-routingSkill
- lean-on-oss-standardsSkill
- live-money-disciplineSkill
- live-state-verificationSkill
- loop-governance-and-learningSkill
- low-level-executor-task-specSkill
- map-dispatch-over-conditionalsSkill
- merge-conflict-disciplineSkill
- mermaid-diagramsSkill
- oke-arc-runners-opentofuSkill
- orchestrate-agySkill
- orchestrate-claudeSkill
- orchestrate-codexSkill
- orchestrate-geminiSkill
- otel-emit-at-willSkill
- parameter-designSkill
- parse-dont-validateSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Verify vantage before diagnosing access.
- Distinctive: line 16
Check authenticated account, repo, PR URL/number, base branch, and whether missing data is an auth problem or a real PR state.
- Distinctive: line 17
Read PR state from GitHub fields, not local guesses.
- Distinctive: line 32
Do not push just to retrigger checks unless there is evidence the old run is stale and a legitimate retrigger path is unavailable.
- Distinctive: line 33
Do not churn red CI. Fix failures locally, then push one focused update.
- Distinctive: line 37
Leave blocker comments only when useful and evidenced.
- Distinctive: line 39
Keep comments concise and omit sensitive details, credentials, private logs, or speculative blame.
- Very specific: line 49
If the PR cannot be merged, report blocked with the classification, the GitHub evidence observed, the exact missing requirement, and the …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
GitHub PR Shepherding
Overview
Shepherd a pull request from current GitHub state to either merged or explicitly blocked. Treat GitHub as the system of record, avoid CI churn, and make every wait, comment, or unblock action evidence-based.
Workflow
- Verify vantage before diagnosing access.
- For local GitHub CLI reads, run
env -u GH_TOKEN -u GITHUB_TOKEN gh ...so environment tokens do not shadow host auth. - Check authenticated account, repo, PR URL/number, base branch, and whether missing data is an auth problem or a real PR state.
- For local GitHub CLI reads, run
- Read PR state from GitHub fields, not local guesses.
- Detect native stack membership before choosing merge commands. Follow
stacked-diff-disciplinefor the ordered members, trunk, selected prefix, and per-member heads/bases/checks. Distinguish a native stack from ordinary parent-based PRs. - Inspect draft state, merge state/status, review decision, auto-merge request, merge queue state, base branch, head branch, and
statusCheckRollup. - For checks, capture the exact check names, conclusions/statuses, URLs, and last update times needed to justify the classification.
- Detect native stack membership before choosing merge commands. Follow
- Classify the PR into exactly one current state:
merged: merged into the intended base.mergeable: approved, non-draft, checks satisfied, and the applicable PR or stack merge can proceed. A native member also requires every unmerged member below it to be ready.waiting on checks: checks are pending or queued with recent activity.failed checks: required or blocking checks failed or were cancelled.missing review/signoff: review decision or required approval is blocking.draft: draft state blocks merge or queue entry.queued/no runner pickup: queued, requested, or expected CI work has not started; distinguish normal queue delay from stale pickup.inaccessible due to auth: GitHub data needed for classification cannot be read after vantage is verified.
- Choose the least disruptive legitimate action.
- For an ordinary PR, enable or preserve supported auto-merge when requirements are expected to clear. Native stacks do not support auto-merge: use
gh stack merge <highest-included-pr> --yesor its native queue path after verifying the selected prefix. Do not use a per-PR auto-merge loop for a native stack. - Do not push just to retrigger checks unless there is evidence the old run is stale and a legitimate retrigger path is unavailable.
- Do not churn red CI. Fix failures locally, then push one focused update.
- Bound CI using actual workflow and runner controls. Drafts suppress jobs only when workflows implement that behavior. Promote all members of a ready merge prefix within capacity; parent merge is not a universal prerequisite for child review or CI.
- A lower-head, trunk, order, or membership change requires fresh applicable evidence. For a queued stack's normal trunk advancement, verify the queue's current merge-group SHA/checks without competing branch rewrites. Queue admission remains
queueduntil GitHub confirms actual merges. - For long waits, checkpoint with timestamped evidence and the next planned recheck instead of repeatedly polling without new information.
- For an ordinary PR, enable or preserve supported auto-merge when requirements are expected to clear. Native stacks do not support auto-merge: use
- Leave blocker comments only when useful and evidenced.
- State the exact blocker, observed GitHub status, and next legitimate unblock path.
- Keep comments concise and omit sensitive details, credentials, private logs, or speculative blame.
Done Criteria
Done means the PR is merged into the intended base and the real artifact is verified from the default branch or other system of record the user cares about. For bug fixes, verifier fixes, or CI fixes, include same-PR regression or ratchet evidence before calling it done.
For a stack, verify every included PR and map it to the resulting trunk SHA or queue group and post-merge evidence. Verify the combined artifact and remaining members' new bases/heads; one green member does not establish stack delivery.
If the PR cannot be merged, report blocked with the classification, the GitHub evidence observed, the exact missing requirement, and the next legitimate owner/action.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

