AgentSkillsCN

lgtm

展示知识库的最新状态与近期学习成果。

SKILL.md
--- frontmatter
name: lgtm
description: Launch lgtm TUI to review a markdown file with line-by-line commenting
argument-hint: <file-path> [--stdin]
allowed-tools: Bash(lgtm *), Bash(tmux *), Read

Launch lgtm to review the specified markdown file or piped content.

Usage

/lgtm path/to/plan.md # Opens TUI for review (auto-spawns in tmux pane) /lgtm @CLAUDE.md # @ prefix is stripped automatically /lgtm --stdin # Review content piped from stdin

What it does

Opens an interactive TUI for reviewing markdown documents with:

  • •Line-by-line navigation (j/k)
  • •Comment types: Blocker, Concern, Question, Suggestion, Praise, Acknowledge
  • •Syntax highlighting for code blocks
  • •Export to markdown (E) or clipboard (y)
  • •Summary view (v)

Instructions

When this skill is invoked, follow these steps:

  1. •Parse arguments: Extract the file path from $ARGUMENTS. If it starts with @, strip that prefix. Ignore --tmux if present (it's automatic).
  2. •Resolve path: Convert the file path to an absolute path if it isn't already.
  3. •Read the file: Use the Read tool to load the full file content into context. This is essential so you can correlate the user's line-level comments with actual file content.
  4. •Create export path: Generate /tmp/lgtm-export-{basename-without-extension}.md
  5. •Run the TUI: Execute the command below. The --export-on-quit flag automatically enables tmux mode and waits for the user to quit.
    bash
    lgtm "<resolved-file-path>" --export-on-quit "<export-path>"
  6. •Read the export: After the command returns (user has quit the TUI), use the Read tool to read the export file.
  7. •Respond in review mode: This is a collaborative review workflow. After reading the export:
    • •Correlate each comment with the corresponding line(s) from the original file content you read in step 3.
    • •Summarize the user's feedback, grouping by comment type (Blockers, Concerns, Questions, Suggestions, Praise).
    • •Proactively address blockers and concerns first — propose concrete code or content changes to resolve them.
    • •Answer any questions the user raised.
    • •Acknowledge suggestions and praise.
    • •Offer to iterate on the file by applying the proposed changes.