# qqling-skills
[English](README.md) | [简体中文](README.zh-CN.md)
A collection of self-contained Agent Skills for Codex, Claude Code, and other agents that support the Skills format, covering repository diagnostics, Skill usage analysis, shift scheduling, photo fusion, and WeChat mini program launch materials.
Each Skill includes a `SKILL.md` entry point and its own scripts and references. Available capabilities depend on the tools and data accessible to the host agent; `ql-skill-usage-auditor` currently analyzes Codex session logs.
## Skills
| Name | Description |
| --- | --- |
| [ql-reality-fusion](ql-reality-fusion/SKILL.md) | Blends photography with 16 styles, including anime, comic halftones, layered paper, pixels, voxels, felt, ink wash, and clay. Edits selected people, backgrounds, or local elements while preserving identity, composition, lighting, and occlusion. Requires an image editing tool that accepts reference images. |
| [ql-harness-doctor](ql-harness-doctor/SKILL.md) | Audits a single Git repository against 46 Harness controls consolidated from 14 research chapters. Reports six diagnostic states, control coverage, L0–L4 maturity, and grouped recommendations. Checks task status transitions and recovery after interruptions, and distinguishes declared rules, execution records, and verified results. Runs only when explicitly invoked by name. |
| [ql-skill-usage-auditor](ql-skill-usage-auditor/SKILL.md) | Counts actual Codex Skill loads and provides rankings by period, trend visualizations, coverage status, and safe cleanup recommendations. Skill names or `$skill-name` mentions in ordinary text do not count as usage. |
| [ql-wechat-game-launch-kit](ql-wechat-game-launch-kit/SKILL.md) | Takes a WeChat mini program or mini game from a fresh account to release. Scans the codebase and writes paste-ready materials, one folder per backend form: a registration folder (account, subject type, names and short names, logo candidates, intro, service category, operating region, WeChat verification, age rating, qualification path, privacy guide, ICP filing), the mini game content filing (scenes, gameplay, systems, screenshots), and review/release (server domains, version notes). Checks package size, network domains and privacy-consent APIs, verifies that no screenshot repeats across modules, and walks the user through each form step by step with saved progress. Does not submit anything on the platform. |
| [ql-schedule-night-shifts](ql-schedule-night-shifts/SKILL.md) | Creates departmental night-shift schedules in Excel from staff availability, the target month, leave, and individual requirements. Supports Chinese public holidays, a fairness ledger across months, candidate schedule approval, and minimal-change repairs after unexpected leave. |
## Installation
Install the Skill for your agent, such as Codex or Claude Code. Replace `<skill-name>` with one of the `ql-` Skill names listed above:
```bash
npx skills add qshine/qqling-skills --skill <skill-name>
```
## Reality Fusion
Upload a photo and ask, for example:
> Use $ql-reality-fusion to turn the person into an anime character while keeping the landscape photographic. Keep the pose and clothing, with natural lighting and contact shadows.
For an open request, the Skill recommends suitable combinations of edit region and style. Explicit instructions or authorization to choose the best option proceed directly to editing. Host tool interaction rules take precedence. The default uses bold stylization so the photographic and virtual regions are immediately distinguishable in the whole image. Natural integration comes from coherent light, occlusion and contact, while the rendering media remain distinct. Explicit requests for subtle edits or small targets are respected. One finished image is delivered, with up to two targeted corrections for visible problems.
Generation requires a host image editing tool with reference-image support; the Python helper only validates a grounded brief and composes an edit prompt. Codex's built-in image tool is the preferred route. Python 3.9+ is only needed for the optional helper, which has no package dependencies. This Skill does not promise pixel-identical preservation or validated quality for every photo/style combination.
From the repository root, validate the Skill and run its tests:
```bash
python ql-skill-usage-auditor/scripts/validate_skill.py ql-reality-fusion
python -B -m unittest discover -s ql-reality-fusion/tests -p 'test_*.py' -v
```
Behavior and visual acceptance scenarios are in [the evaluation cases](ql-reality-fusion/tests/evaluation-cases.md). Keep personal photos and generated images out of version control.