スキル一覧に戻る
lnittman

skill-publish

by lnittman

Claude Code skills for power users

0🍴 0📅 2026年1月23日
GitHubで見るManusで実行

SKILL.md


name: skill-publish description: One-way sync from private skills to public versions. Triggers include "publish skills", "sync to public", "update public skills", or when private skills have changed.

skill-publish

one-way transformation of private skills to portable public versions. strips private CLI dependencies, documents alternatives.

philosophy

"the decision tree is the skill; the CLI is convenience"

principleapplication
one-way flowprivate → public only, never reverse
preserve decision treescore methodology stays intact
document alternativesprivate CLIs become optional enhancements
exclude personalcommunication and orchestration skills stay private
minimal diffonly change what's necessary for portability

when to use

useskip
private skill changed significantlycosmetic edits only
new skill ready for sharingskill still in development
periodic sync (monthly)just synced recently
team member needs updated skillskill is in excluded list
preparing for Claude Desktop uploadskill depends on local secrets

decision tree: include or exclude

Should this skill be published?
├── Personal communication skill? (imessage, slack)
│   └── EXCLUDE: contains personal voice/patterns
├── Heavy orchestration skill without portable bootstrap? (auto)
│   └── EXCLUDE: requires local agent infrastructure
├── External AI collaboration/orchestration skill? (pair, fanout, loop)
│   └── INCLUDE: note bash + copilot/codex requirements
├── Has private CLI as core dependency?
│   ├── CLI is enhancement only? → INCLUDE with alternatives
│   └── CLI is essential to function? → EXCLUDE
├── Contains secrets or API patterns?
│   └── EXCLUDE or redact
└── Default
    └── INCLUDE with transformation

decision tree: transformation scope

How much transformation needed?
├── No private CLI references?
│   └── copy as-is (rare)
├── Only ## tool integration section?
│   └── light: convert to ## enhancements (optional)
├── CLI commands in workflow steps?
│   └── medium: add manual alternatives inline
├── CLI commands throughout?
│   └── heavy: systematic replacement + enhancements section
└── CLI is core to skill function?
    └── exclude from public

decision tree: CLI categorization

What kind of CLI is this?
├── Private/local only?
│   ├── outline, layer, verify → manual code exploration
│   ├── trails → manual notes or omit
│   ├── linear → generic issue tracker patterns
│   ├── slack → generic messaging patterns
│   ├── agents, codex, copilot → omit or generalize
│   ├── prompts → inline the content or omit
│   └── messages → omit (personal)
├── Public CLI (available via package managers)?
│   ├── gh (GitHub CLI) → keep, widely available
│   ├── rg (ripgrep) → keep with grep alternative
│   ├── fd (fd-find) → keep with find alternative
│   └── jq → keep, widely available
├── Standard Unix?
│   └── keep: wc, grep, find, git, cat, stat, readlink
└── Unknown?
    └── check if installable, else provide alternative

concrete values

itemvaluesource
included skills18current public/ inventory
excluded skills3imessage, slack, auto
sync frequencymonthly or on significant changeconvention
enhancements sectionalways last before referencesstructure convention
alternative formattool | command | alternative tablereadability

included skills

skillcategoryprimary value
ask-deepclarificationOARS questioning technique
emil-kowalskidesignanimation principles
pr-auditreviewsecurity/performance checklist
test-pilottestingcoverage strategies
friction-analysisanalysisDX bottleneck detection
contextcontinuityhandoff patterns
agent-docs-auditdocsdocumentation accuracy
project-reviewreviewstandards alignment
issue-contextenrichmentissue preparation
skill-composemetaskill orchestration
www-studiodesignmarketing site craft
metaprompt-factorymetaprompt engineering
skill-auditmetaskill inventory
skill-createmetaskill scaffolding
skill-improvemetaskill enhancement
paircollaborationoutput contracts for external AI
fanoutanalysismulti-perspective parallel analysis
looporchestrationautonomous work with zero-context bootstrap

excluded skills

skillreason
imessagepersonal communication patterns
slackworkspace-specific, requires tokens
autorequires local agent infrastructure (no portable bootstrap)

workflow

phase 1: inventory

# list private skills
ls ~/Developer/skills/*/SKILL.md

# list current public skills
ls ~/Developer/skills/public/*/SKILL.md

# find skills needing sync (modified after public version)
for skill in ~/Developer/skills/*/SKILL.md; do
  name=$(basename $(dirname "$skill"))
  # skip public dir itself
  [ "$name" = "public" ] && continue
  public="$HOME/Developer/skills/public/${name}/SKILL.md"
  if [ -f "$public" ]; then
    if [ "$skill" -nt "$public" ]; then
      echo "needs sync: $name"
    fi
  fi
done

phase 2: assess each skill

for each skill needing sync:

  1. check exclusion list
  2. identify CLI dependencies
  3. determine transformation scope
  4. note sections requiring changes

phase 3: transform

transformation checklist:

  • update frontmatter description (remove private tool mentions)
  • replace private CLI commands with manual alternatives
  • convert ## tool integration to ## enhancements (optional)
  • add alternatives table for each private CLI
  • preserve all decision trees unchanged
  • preserve all concrete values unchanged
  • preserve all anti-patterns unchanged

phase 4: write public version

# create skill folder
mkdir -p ~/Developer/skills/public/{skill-name}

# write SKILL.md (proper skill structure)
# ~/Developer/skills/public/{skill-name}/SKILL.md

phase 5: verify

verification checklist:

  • decision trees intact and actionable
  • no broken references to private tools
  • enhancements section has alternatives for each tool
  • frontmatter name matches filename
  • no personal/private patterns leaked

transformation patterns

tool integration → enhancements

before (private):

## tool integration

| tool | command | purpose |
|------|---------|---------|
| layer | `layer .` | understand architecture |
| outline | `outline --callers=X` | trace dependencies |
| trails | `trails trail record` | persist context |

after (public):

## enhancements (optional)

These CLI tools provide automation but are not required:

| tool | command | alternative |
|------|---------|-------------|
| layer | `layer .` | Read package.json files, trace imports manually |
| outline | `outline --callers=X` | Use grep to find function references |
| trails | `trails trail record` | Manual notes |

inline CLI references

before:

### phase 1: explore
```bash
layer .                    # architecture
outline --callers=X src/   # dependencies

after:
```markdown
### phase 1: explore
```bash
# understand architecture
# - read package.json for dependencies
# - check directory structure

# trace dependencies
# - grep for function/import references
# - or use IDE "find references"

### workflow commands

before:
```markdown
```bash
verify --format=summary    # run tests
linear issue view ARB-123  # get context

after:
```markdown
```bash
# run tests using your test runner
npm test
# or: pnpm test, yarn test, pytest, etc.

# get issue context from your issue tracker
# check the issue description and comments

## anti-patterns

| pattern | problem | fix |
|---------|---------|-----|
| reverse sync | public changes lost | always private → public |
| include non-portable orchestration | broken without infrastructure | exclude auto (has no portable bootstrap) |
| strip decision trees | removes core value | preserve all decision trees |
| vague alternatives | "use something else" | specific alternative commands |
| include secrets | security risk | redact or exclude |
| sync too often | churn | monthly or significant changes |
| forget enhancements | users can't enhance | always add section |

## quality checklist

before committing public version:

- [ ] all decision trees preserved
- [ ] all concrete values preserved
- [ ] all anti-patterns preserved
- [ ] no private CLI commands without alternatives
- [ ] enhancements section present (if any tools referenced)
- [ ] frontmatter description updated
- [ ] no personal patterns (voice, accounts, secrets)
- [ ] skill is self-contained and actionable

## references

- [../public/README.md](../public/README.md) - public skills documentation

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です