wangyitong a65adcc2e5 Initial commit: merged, deduplicated, and vetted skill collection

Sources: extracted from two upstream archives (skill-repo, skills-main),
merged with the following policy:

- 15 broken symlinks (pointing to /Users/jameslee/.cc-switch/skills or
  ../../.agents/skills on a foreign machine) discarded
- 3 real name collisions with identical content (ai-pair, ifind-http-api,
  zhipu-websearch) kept as one copy
- Functional overlaps deduped keeping the strongest variant:
  - docx family: kept docx (official, full toolchain) + docx-cn
    (GB/T 9704 Chinese official-document constants),
    dropped docx_writer (no scripts, name collided with docx)
  - humanizer family: kept humanizer-zh (6 zh reference docs),
    dropped humanizer (en, redundant for CN workflow)
- Skills that only ran in a foreign environment removed:
  ablemind-ops, app-publish, hlb-design-system, openclaw-adj-skill,
  claude-driver
- alphapai excluded from this public repo because its SKILL.md hard-coded
  live credentials

Result: 25 skills, 572 files, ~7.5 MB.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-07-13 14:47:12 +08:00

2.9 KiB

Raw Blame History

Content Team Example

You've drafted a newsletter about AI memory systems and want multi-perspective feedback before publishing.

Start the team

/ai-pair content-team AI-Newsletter

Output:

Team ready.

Team: AI-Newsletter-content
Type: Content Team
Members:
  - author: ready
  - codex-reviewer: ready
  - gemini-reviewer: ready

Awaiting your first task.

Assign a task

Review this draft newsletter article about why AI memory
(like MEMORY.md in Claude Code) isn't actually "memory" in the
way most people think. The article argues that it's closer to
a post-it note than a brain.

[paste your draft or provide file path]

The author agent reads the draft and prepares it for review. You approve sending to reviewers.

Review results

## Codex Review

### Logic & Accuracy
- The claim "LLMs have no persistent state between sessions" is technically
  correct but oversimplified. Some systems do maintain vector stores that
  persist. Consider qualifying with "by default" or "natively".
- The analogy to post-it notes breaks down: post-it notes don't have
  retrieval logic. MEMORY.md is more like a briefing document.

### Fact-Checking
- The reference to "context window limits of 200K tokens" should specify
  this is Claude's limit. GPT-4 Turbo is 128K.

### Summary
Solid argument, but two factual claims need tightening.

## Gemini Review

### Readability & Flow
- The opening paragraph uses three technical terms before the reader
  gets the hook. Consider leading with the relatable scenario first.

### Engagement & Hook
- The "post-it note" metaphor is strong but introduced too late (paragraph 4).
  Move it to the opening for immediate impact.

### Audience Fit
- Tone shifts between casual (paragraphs 1-3) and academic (paragraphs 4-6).
  The casual tone matches the target audience better — maintain it throughout.

### Summary
Good content, needs structural reorganization for maximum impact.

Codex caught factual precision issues. Gemini caught readability and structure issues. Zero overlap.

Iterate

You tell Team Lead:

Fix the factual claims Codex flagged.
Move the post-it metaphor to the opening as Gemini suggested.
Keep the casual tone throughout.
Don't change the core argument.

The author revises. You can send for another round of review or pass.

End the session

/ai-pair team-stop

Tips for Content Team

Provide context about your audience — reviewers give better feedback when they know who's reading
Don't fix everything — you decide which feedback matters. Codex tends to over-index on precision; Gemini tends to over-index on accessibility
Use iteratively — first round for big issues, second round for polish
Style memory — if you have a style-memory.md file, the author agent will automatically follow your style preferences

2.9 KiB Raw Blame History