SocketDev/action

cascading-fleet

Propagate a wheelhouse template change across fleet repos: worktrees, push/PR fallback, cleanup.

agent-ci

Run this repo's GitHub Actions locally with Agent-CI before pushing CI-sensitive changes.

auditing-api-surface

Audit package exports for dead, internal-only, or weakly-consumed subpaths before pruning.

auditing-gha

Audit Actions permissions/allowlists against the fleet baseline; --conform fixes drift.

authoring-spec

Turn the current conversation into a spec: problem, solution, stories, tests, enforcement.

building-tdd

Use red-green-refactor for a feature or bug fix when the user asks for test-first work.

cleaning-ci

Remove redundant CI workflows, stale Dependabot auto-fix surface, and obsolete workflow run clutter.

codifying-disciplines

Turn repo rules that live only in prose or memory into scripts, hooks, lint rules, or checks.

consolidating-commits

Regroup commits since a base into logical commits, keeping any version bump last; not squash-to-one.

creating-guards

Author or convert fleet hooks to the uniform check plus runGuard contract used by the dispatcher.

decomposing-tickets

Break a plan, spec, or findings report into independently-grabbable vertical-slice tracker tickets.

deduping-dependencies

Reduce duplicate installs with overrides, hardened drop-ins, patches, consumer checks.

delegating-execution

Route substantial work through plan, execute, review, follow-up agents at the right effort tier.

designing-interfaces

Designs and reviews accessible interfaces using reference research and rendered validation.

diagnosing-bugs

Run the diagnosis loop for hard bugs, failing behavior, regressions, thrown errors, or slow paths.

driving-cursor-bugbot

Review Cursor Bugbot PR threads, classify findings, fix real bugs, reply inline, and push updates.

extracting-design-systems

Consolidates repeated UI patterns. Use after a pattern proves stable.

gh-stack

Runs gh stack for dependent PRs and preview feedback.

greening-ci-local

Run Agent-CI locally in Docker, fix paused failures, and retry until the workflow is green.

greening-ci

Watch remote CI, pull the first failure log, fix locally, push, and loop until the run is green.

grilling-plan

Stress-test a plan one question at a time before building, especially for shared fleet resources.

grooming-backlog

Collapse noisy backlogs: verify done work, fold duplicates, delete stale tasks.

guarding-paths

Enforce one constructed path per concern; fix duplicated build, test, runtime, config paths.

handing-off

Write a compact handoff so another agent or human can continue the work without replaying context.

improve

Read-only codebase survey returning prioritized implementation plans for other agents.

improving-web-interfaces

Improves web interfaces after design direction is locked.

locking-down-claude

Secure non-interactive Claude/SDK calls: pinned tools, prompts, permissions.

looping-quality

Run scanning-quality, fix findings, repeat until clean or the iteration limit is hit.

managing-pnpm-workspaces

Maintains pnpm workspace dependencies safely.

managing-worktrees

Create, fan out, and prune git worktrees for isolated tasks and parallel PR review.

map

Emit a token-cheap symbol map for files or directories before reading unfamiliar source in detail.

measuring-ecosystem-impact

Rank npm packages by ecosystem reach; model what overriding them removes from the install tree.

migrating-rule-packs

Run deterministic 10+ file code migrations via rule-pack transforms and isolated agents.

opening-pr

Implement a requested change, verify it, commit it, and open or update the PR.

optimizing-compiler-performance

Audits compiler and codegen performance.

optimizing-cpp-performance

Audits measured C++ performance.

optimizing-go-performance

Audits measured Go performance.

optimizing-javascript-performance

Audits measured JavaScript and TypeScript performance.

optimizing-memory-performance

Audits measured memory and allocation performance.

optimizing-node-native-performance

Audits Node addon and native-boundary performance.