Qi Layer
Write and maintain AGENTS.md and .context/CONTEXT.md as layered codebase docs
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 31, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · e47a903
⚠ The author changed this skill after we tested it. The verdict below describes the version we ran on Jul 31, 2026; it's queued for a re-test.
Fetched SKILL.md (HTTP 200) and resources/references.md (200); both referenced sibling skills knowledge-layers and llm-writing resolve (200). Frontmatter parses with name+description; no scripts, no curl|sh, no exfiltration — references.md only cites two external URLs as design sources. Output test: wrote an AGENTS.md for a sample sliding-window rate-limit module both without and with the skill. Baseline narrated the code file-by-file and carried a "TODO"/"in-memory" status section (exactly the session-bleed the skill flags); the skill version instead led with decision-critical footguns the baseline never stated ("effective global limit is max_hits × workers", "peek consumes two slots", "a False means slow down, not try again") and moved algorithm/Redis/O(n) detail into .context/CONTEXT.md. Trigger judgments: 3 authoring/trimming phrasings load it, docstring and README-install phrasings correctly do not (5/5).
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 7/10
- Docs & honesty 5/5
What Qi Layer does
A guidance skill for authoring the code-local documentation pair AGENTS.md and .context/CONTEXT.md, using four principles (fractal compression, hierarchical summarization, LCA deduplication, progressive disclosure) plus explicit tests for what belongs at each level. Triggers when you are writing, trimming, or maintaining these agent-facing intent docs for a codebase — not for docstrings, READMEs, or code-level comments. It pushes reference detail down into .context/ and keeps root AGENTS.md focused on decision-critical knowledge.
How to install Qi Layer
git clone --depth 1 https://github.com/haowjy/creative-writing-skills.git /tmp/qi-layer-src
mkdir -p ~/.claude/skills
cp -R /tmp/qi-layer-src/cw/skills/qi-layer ~/.claude/skills/qi-layer
# SKILL.md lands at ~/.claude/skills/qi-layer/SKILL.md ; bundled resources/references.md comes with it.
# Soft deps: SKILL.md body says "Load /knowledge-layers" and "/llm-writing". Those are sibling
# skills in the same repo (cw/skills/knowledge-layers, cw/skills/llm-writing). The core guidance
# is fully inline and works standalone; install those two the same way if you want the extras.
# No API keys, no build step, no scripts to run.
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Qi Layer
-
/qi-layerWrite and maintain AGENTS.md and .context/CONTEXT.md as layered codebase docs
It also activates on plain-language prompts like these:
-
Explore this codebase using AGENTS.md as context -
Update CONTEXT.md after this change -
Keep this intent doc concise before committing
Frequently asked questions
- Is the Qi Layer skill free?
- Yes. The skill itself is free from haowjy/creative-writing-skills. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Qi Layer work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 31, 2026. Verdict: Tested · Works. Fetched SKILL.md (HTTP 200) and resources/references.md (200); both referenced sibling skills knowledge-layers and llm-writing resolve (200). Frontmatter parses with name+description; no scripts, no curl|sh, no exfiltration — references.md only cites two external URLs as design sources. Output test: wrote an AGENTS.md for a sample sliding-window rate-limit module both without and with the skill. Baseline narrated the code file-by-file and carried a "TODO"/"in-memory" status section (exactly the session-bleed the skill flags); the skill version instead led with decision-critical footguns the baseline never stated ("effective global limit is max_hits × workers", "peek consumes two slots", "a False means slow down, not try again") and moved algorithm/Redis/O(n) detail into .context/CONTEXT.md. Trigger judgments: 3 authoring/trimming phrasings load it, docstring and README-install phrasings correctly do not (5/5).
- What is the Qi Layer SkillProof Score?
- 8.8/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 7/10, docs & honesty 5/5.
- How do I install Qi Layer?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Qi Layer with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.