sipengxie2024
24 verified skills24 total stars
executing-plans
Use when executing a written implementation plan in a separate or manual session with batch checkpoints and .planning updates.
testing1
planning-foundation
Use when starting complex multi-step tasks, research projects, or anything needing >5 tool calls. Provides the persistent .planning/ directory used as on-disk working memory across sessions.
tools1
verification
Use when about to claim work is complete, fixed, or passing, before committing or creating PRs.
data-ai1
tdd
Use when implementing any feature or bugfix, before writing implementation code
development1
lightweight-execute
Use when task has clear requirements, touches 2-5 files, and needs structured tracking without full brainstorming/planning/review pipeline
devops1
stashing
Use when switching to another project, waiting on dependencies, or temporarily setting aside unfinished work.
testing1
main
Skill router and planning initialization. Loaded on every session start. Determines which skills to invoke and ensures .planning/ is initialized for complex tasks.
data-ai1
spec-interview
Use when refining design docs and specs through deep questioning. Triggered when user says "help me refine this plan", "ask me about the project", "deep interview", or after brainstorming produces a design doc.
documentation1
collaborating-with-codex
Use when delegating coding work — prototyping, debugging, analysis, review, implementation, or generating publication-quality scientific/architecture diagrams — to Codex CLI. Supports multi-turn sessions via SESSION_ID.
tools1
debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
development1
writing-plans
Use when a spec or requirements exist for a multi-step task and an implementation plan needs to be written before touching code
development1
subagent-driven-codex
Use when executing an implementation plan in this session, sequentially, by routing implementer and reviewer roles through Codex CLI. Two-stage Codex review per task.
tools1
receiving-review
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable.
development1
session-handoff-audit
Use when wrapping up a long multi-commit session, handing off to a new session, or the user asks to "audit what we've done", "verify implementation state", "check for drift", "防止幻觉", or "跨会话交接". Generates a self-contained prompt that a fresh, context-free session uses to independently verify the work against design specs. Generates the prompt only — does not perform the audit itself.
testing1
releasing
Use when publishing a new version of the plugin.
tools1
subagent-driven
Use when executing an implementation plan in this session, sequentially, dispatching one Claude subagent per task with two-stage review.
data-ai1
team-driven
Use when executing implementation plans with parallel task groups or individual tasks too heavy for subagent context limits.
data-ai1
archiving
Use after completing a plan or when .planning/ has accumulated stale data and needs a clean reset for the next task.
data-ai1
finishing-branch
Use when implementation is complete, all tests pass, and integration strategy needs to be decided - merge, PR, keep, or discard
testing1
requesting-review
Use when completing plan-driven tasks, implementing major features, or before merging to verify work meets requirements.
testing1
parallel-agents
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
testing1
git-worktrees
Use when starting feature work that needs isolation from the current workspace or before executing implementation plans.
tools1
brainstorming
Use when designing complex multi-step features, building new components, or planning significant behavior changes that require design exploration before implementation.
development1
dual-review
Use when the user asks for a "dual review", "double check", "second opinion", "review before commit/merge", or wants two independent reads on a code change. Runs an internal simplify lens and Codex in parallel review-only, then gates every edit on explicit user approval.
development1