Huiyan Wan
Community@wan-huiyan · Windsor, UK
Huiyan Wan maintains a large registry of diagnostic skills for git worktree isolation, parallel pull-request collisions, squash-merge traps, and multi-agent orchestration failure recovery.
Agent Skills by Huiyan Wan
Showing 121 vetted skills indexed across 3 GitHub repositories.
ab-harness
Measures whether a Claude Code setup improves task outcomes via counterfactual A/B and layered-ablation runs.
green-gate-never-examined-the-case
Diagnoses CI gates and tests that report green without examining the case they guard.
skill-portfolio-repo-placement-scan
Maps authored Claude Code skills to target repos with ADD, UPDATE, and CROSS-LINK recommendations.
test-effectiveness-auditor
Replays historical bugs at pre-fix commits to measure whether test suites catch real defects.
claude-plugin-repo-ci-release
Adds CI validation and release-on-version-bump GitHub Actions to Claude Code plugin repos.
ecosystem-audit
Audit Claude Code ecosystems by scanning skills, memory, handoffs, ADRs, and worktrees into a scored HTML report.
memory-hygiene
Audits and cleans Claude Code memory files, lessons, ADRs, and docs taxonomy.
doc-freshness-reverse-lint
Detect stale normative guidance in project docs after memory-file rules change.
repo-hygiene
Audits repositories before pull requests for committed data files, hardcoded paths, and tracked artifacts.
context-police
Audit, curate, and measure the token cost of an agent harness's always-on skills catalog.
stale-base-pr-silently-reverts-upstream-content
Detect and recover content silently reverted by stale-base sibling PR merges.
parallel-subagent-fanout-rate-limit-recover-from-disk
Recovers throttled parallel subagent fan-outs by checking disk output instead of agent return status.
using-git-worktrees
Creates isolated git worktrees with directory selection, gitignore verification, and baseline test validation.
verify-live-prod-before-shipping-superseded-fix
Verifies live production state before deploying fixes that parallel sessions may have superseded.
stacked-pr-base-branch-deletion-auto-closes-dependent
Recover stacked GitHub PRs auto-closed by base-branch deletion after squash merge.
auto-merge-rearms-while-agent-live-kill-then-disarm-verify
Holds GitHub pull requests by killing the agent, disarming auto-merge, and verifying state.
parallel-pr-template-fork-duplicates-moved-section
Diagnose silent template duplication after parallel PRs move and fork the same section.
task-framing-claims-need-subagent-grep-verify
Instructs write-content sub-agents to grep-verify and tag corrections to dispatcher task-framing claims.
gh-squash-merge-closes-only-one-issue
Diagnoses and fixes GitHub squash-merge auto-close leaving referenced issues open.
parallel-impl-agent-dies-mid-stream-verify-working-tree
Verifies actual working-tree output when a parallel subagent dies mid-stream despite reporting completion.
git-stash-pop-pulls-unrelated-stash
Diagnose and prevent git stash pop from applying unrelated stashes after a no-op stash.
worktree-stale-local-main-ref-inflates-pr-diff
Diagnoses inflated git diff output caused by a stale local main ref in multi-worktree repositories.
cross-worktree-spec-handoff-via-checkout-paths
Transfers committed files between parallel git worktrees using checkout path restoration.
pr-conflict-from-mid-flight-merges
Diagnose and resolve GitHub PR conflicts caused by sibling PRs landing on main mid-flight.
Frequently Asked Questions About Huiyan Wan
FAQPage SchemaWhat tasks can I accomplish with Huiyan Wan's skill registry?▼
You can diagnose and recover from git worktree conflicts, stale-base PR reversions, stranded squash-merge commits, async post-commit hook corruption, subagent stalls and rate limits, GitHub Actions deploy-gate misconfigurations, and session handoff collisions in multi-session repositories.
Who are these skills designed for?▼
Engineers running parallel Claude Code sessions, multi-agent orchestrations, and dense concurrent PR traffic in shared repositories — particularly teams using git worktrees, squash-merge flows, stacked PRs, and CI-gated auto-deploy pipelines on GitHub.
How do the skills work in practice?▼
Each skill is a markdown frontmatter document with trigger conditions, root-cause analysis, and exact recovery commands (gh, git, grep). Most set disable-model-invocation: true, meaning they load as reference runbooks invoked explicitly rather than auto-triggered by the model.
What prerequisites do these skills assume?▼
A GitHub-centric environment with the gh CLI, git worktrees (typically under .claude/worktrees/), and frequently Claude Code's Agent/Task/Workflow dispatch mechanisms. Several skills also assume BigQuery, gcloud, or GitHub Actions deploy pipelines where noted.
Are these skills free to use?▼
The manifest lists them as public repository content authored by Claude Code and wan-huiyan with no stated license or cost terms; they are plain markdown skill definitions readable and adaptable without any runtime dependency or paid component.