Commands and Workflow Reference

Use this page as a quick contract: entry, purpose, and output.

Workflow Skills

LingXi workflow is skill-driven: task -> vet -> plan -> build -> review.
These skills run only when explicitly invoked (command or clear natural-language request).

Skill	Purpose	Typical output
task	Define goal, boundaries, acceptance	*.task.*.md
vet	Review task quality	In-chat review conclusions
plan	Build implementation and test mapping	*.plan.*.md + *.testcase.*.md
build	Implement and test	Code and test changes
review	Requirement-level acceptance audit	*.review.*.md

Memory Commands

/remember

/remember <description>

• Purpose: capture one memory immediately
• Output: write to memory/project/ or memory/share/, update INDEX

/memory-govern

/memory-govern [--dry-run] [--skip-govern] [--root <memoryRoot>]

• Purpose: index sync and optional proactive governance
• Effect: restore INDEX/notes consistency and apply governance suggestions after confirmation

/init

/init

• Purpose: initialize project context and generate memory candidates
• Default behavior: candidate-first, no automatic write until explicit confirmation

Debug and Tuning Helpers

about-lingxi (Skill)

• Role: LingXi's human+AI co-readable technical handbook (knowledge base), exposed as a Skill for on-demand automatic loading
• Think of it as one capability with two layers:
  • Documentation layer (Human + AI Readable): background, architecture, mechanisms, design principles, evaluation criteria
  • Invocation layer (Skill): quickly loads that knowledge when the model needs LingXi-specific context
• Use when:
  • you need to align on what LingXi is and how/why it is designed this way
  • you are evaluating long-term benefit/cost/boundary of mechanism changes
  • you are choosing among Command / Skill / Hook / Subagent
• Loading:
  • primarily auto-loaded by the model when LingXi background is needed
  • can also be manually triggered (type /about-lingxi in Cursor and select the skill)
• Boundary: provides background and evaluation context only; performs no concrete actions directly

/start-tuning

/start-tuning

• Role: an upper-layer application of about-lingxi (debug/tuning shortcut entrypoint)
• Goal: switch the model into LingXi tuning mode so developers can debug LingXi efficiently
• What happens: it auto-loads about-lingxi as background context, then starts tuning analysis and discussion
• Use when: debugging LingXi, aligning optimization goals, clarifying constraints, planning evolution paths
• Boundary: session starter for tuning; it does not own the underlying knowledge content (that is about-lingxi)

Automatic Background Tasks (No Manual Command)

At session start, LingXi may trigger background subagents by convention:

• Session distillation via lingxi-session-distill
• Self-iterate via lingxi-self-iterate

Both run in background and do not block the main conversation.

Uninstall

yarn lx:uninstall
# or
npm run lx:uninstall

For non-interactive environments:

yarn lx:uninstall --yes

Install Manifest Automation

Run these three tests in CI:

• install-manifest-exists: manifest path validity (manifest -> repo)
• install-manifest-coverage: structural coverage (repo -> manifest)
• install-manifest-version-sync: version consistency (manifest = package)

Next Steps

• Core Workflow
• Memory System
• FAQ