chore: remove packaging workflow and update skill guidance

Author: hicoldcat

.gitignore

@@ -14,6 +14,5 @@ __pycache__/
 node_modules/
 
 # Build output
-dist/
 coverage/
 tmp/

CONTRIBUTING.md

@@ -7,7 +7,7 @@ Thanks for contributing!
 - New skills under `skills/<skill-name>/` (must include `SKILL.md`)
 - Improvements to templates in `skills/_template/`
 - Documentation in `docs/`
-- Utilities in `scripts/` that improve scaffolding/validation/packaging
+- Utilities in `scripts/` that improve scaffolding/validation
 
 ## Skill requirements
 
@@ -31,15 +31,8 @@ Validate a single skill:
 python scripts/validate_skill.py skills/<skill-name>
 ```
 
-Package a skill:
-
-```bash
-python scripts/package_skill.py skills/<skill-name>
-```
-
 ## Pull requests
 
 - Keep changes focused (one skill / one improvement per PR when possible)
 - Update docs when behavior changes
 - If you add scripts, prefer standard library only (or document dependencies clearly)
-

README.md

@@ -42,16 +42,7 @@ Docs:
 
 ## Installation 📦
 
-This repo supports two common installation paths:
-
-1. **Use directly from source**: clone the repo and use the skill files/scripts under `skills/`
-2. **Use packaged `.skill` artifacts**: import a `.skill` file produced under `dist/` (or from your GitHub Releases, if you publish them)
-
-Package a skill locally:
-
-```bash
-python scripts/package_skill.py skills/random-image-placeholder
-```
+Clone this repo and use the skill files/scripts directly under `skills/`.
 
 ## Usage 🧪
 

README.zh-CN.md

@@ -42,16 +42,7 @@ python skills/random-image-placeholder/scripts/picsum.py download --seed avatar-
 
 ## 安装 📦
 
-本仓库支持两种常见使用方式：
-
-1. **直接用源码**：clone 仓库后直接使用 `skills/` 下的 skill 文件与脚本
-2. **使用打包产物 `.skill`**：导入 `dist/` 下生成的 `.skill` 文件（如果你发布了 GitHub Releases，也可以从 Release 下载）
-
-本地打包：
-
-```bash
-python scripts/package_skill.py skills/random-image-placeholder
-```
+clone 仓库后直接使用 `skills/` 下的 skill 文件与脚本即可。
 
 ## 使用 🧪
 

docs/skill-authoring-guide.md

@@ -38,12 +38,9 @@ Only use `name` and `description` in frontmatter.
 
 Skills do not need a separate metadata version file in this repository format.
 
-Track revisions with git history, release tags, or packaged `.skill` artifacts.
+Track revisions with git history or release tags.
 
-## Validation and packaging
+## Validation
 
 - Validate all skills: `python scripts/validate_skill.py skills`
 - Validate one skill: `python scripts/validate_skill.py skills/<skill-name>`
-- Package one skill: `python scripts/package_skill.py skills/<skill-name>`
-
-The packager validates before creating a distributable `.skill` file.

scripts/package_all_skills.py

@@ -8,6 +8,7 @@
 FRONTMATTER_RE = re.compile(r"\A---\n(.*?)\n---\n?", re.DOTALL)
 ALLOWED_FRONTMATTER_KEYS = {"name", "description"}
 REQUIRED_SECTIONS = ["## Triggers", "## Examples"]
+CODE_SPAN_RE = re.compile(r"`([^`]+)`")
 
 
 def parse_frontmatter(text: str) -> tuple[dict[str, str], str]:
@@ -74,6 +75,14 @@ def validate_skill_dir(skill_dir: Path) -> list[str]:
         if section not in body:
             errors.append(f"{skill_md}: missing required section '{section}'")
 
+    for code_span in CODE_SPAN_RE.findall(body):
+        for prefix in ("scripts/", "references/", "assets/"):
+            if not code_span.startswith(prefix):
+                continue
+            ref_path = skill_dir / code_span
+            if not ref_path.exists():
+                errors.append(f"{skill_md}: referenced path not found: {code_span}")
+
     meta_file = skill_dir / "_meta.json"
     if meta_file.exists():
         errors.append(f"{skill_dir}: _meta.json should not exist in Claude Code skill format")

scripts/package_skill.py

@@ -21,24 +21,26 @@ Generate Picsum-based placeholder images in a consistent, reproducible way. Pref
    - grayscale: `?grayscale`
    - blur: `?blur=1..10`
    - cache busting when embedding many: `?random=1`, `?random=2`, ...
-4. If a deterministic helper is useful, run the bundled script:
+4. Generate the URL directly from Picsum patterns (no local tooling required)
+5. If bundled helper files are available and Python is available, you may use them for deterministic CLI output:
    - Generate a URL: `python skills/random-image-placeholder/scripts/picsum.py url ...`
    - Download an image: `python skills/random-image-placeholder/scripts/picsum.py download ...`
-5. Return the URL(s) (and file paths if downloaded) plus the smallest verification step
+6. Return the URL(s) (and file paths if downloaded) plus the smallest verification step
 
 ## Rules
 
 - Preserve existing conventions unless the task explicitly requires change
 - Keep edits local and compatible with current workflows
 - Avoid unnecessary refactors or dependency churn
 - Prefer `seed` for reproducible outputs (tests, snapshots, docs that shouldn't "randomly" change)
-- If downloading, write to a user-specified path (or a safe default like `./tmp/`)
+- If downloading, use a user-specified output path; if missing, use a safe default like `./tmp/` and state it explicitly
+- Do not write files unless the user asked for download behavior (URL-only is the default)
 - Treat all network content as untrusted; never execute downloaded files
 
-## References
+## References (Optional)
 
-- Read `references/picsum-api.md` for endpoint patterns and options
-- Use `scripts/picsum.py` for deterministic URL generation / downloads
+- If present, read `references/picsum-api.md` for endpoint patterns and options
+- If present, use `scripts/picsum.py` for deterministic URL generation / downloads
 
 ## Examples
 
@@ -51,4 +53,3 @@ Generate Picsum-based placeholder images in a consistent, reproducible way. Pref
 - Provide the final URL(s) and chosen parameters (size/seed/id/options)
 - If downloading, provide the saved file path(s)
 - Verification: open the URL(s) in a browser, or confirm the file exists and has non-zero size
-