mischief-production

Files (22)

• CODEOWNERS · .github/
• dependabot.yml · .github/
• ci.yml · .github/workflows/
• CHANGELOG.md
• LICENSE
• README.md
• SKILL.md
• install.sh
• bulk-repo-import-workflow.md · references/
• claude-code-to-hermes-translation.md · references/
• competitive-analysis-template.md · references/
• competitive-code-reading.md · references/
• external-skill-adaptation-recipe.md · references/
• internal-external-priority-markers.md · references/
• multi-repo-brand-sweep.md · references/
• README.md · skills/mp-deprecated/
• README.md · skills/mp-engineering/
• README.md · skills/mp-in-progress/
• README.md · skills/mp-misc/
• README.md · skills/mp-personal/
• README.md · skills/mp-productivity/
• uninstall.sh

SKILL.md (first lines)

# 顽皮制造 · Mischief Production

Merge multiple external skill sources (GitHub repos, local directories, reference docs) into an existing target skill, producing a unified version that absorbs the best capabilities of all sources while preserving the target's identity.

**When to use**: User provides 2+ external skill sources and wants them merged into an existing skill. Also triggers on: "compare our skills against competitors", "merge these repos into one", "ecosystem audit and upgrade", "对标竞品融合升级".
**When NOT to use**: Creating a skill from scratch (→ `skill-creator`), minor patches (→ `skill_manage patch`), eval/regression testing (→ `skill-eval-gate`).

---

## Core Workflow

### Phase 1 · Audit All Sources

For each source (GitHub repo, local dir, etc.):

1. **Read SKILL.md** — understand the skill's identity, trigger conditions, and core capabilities
2. **Catalog assets** — list all `references/`, `scripts/`, `assets/` files with sizes
3. **Extract unique value** — what does this source have that others don't?

**Output**: A capability matrix comparing all sources across dimensions relevant to the fusion target.

Example matrix:

| Capability | Source A | Source B | Source C | Target (current) |
|------------|----------|----------|----------|-------------------|
| Style library | ✅ 40 styles | ✅ 12 presets | ❌ | ❌ |
| Brand protocol | ✅ 5-step | ❌ | ❌ | ❌ |
| Quality checklist | ❌ | ❌ | ✅ 40 items | ❌ |
| Export formats | PDF only | PDF+PPTX | HTML only | PDF only |

### Phase 2 · Plan the Fusion

Decide the fusion strategy based on the capability matrix:

1. **Identify the backbone** — which source/target provides the primary structure and identity?
2. **Identify absorption targets** — what specific capabilities come from each source?
3. **Check for conflicts** — overlapping capabilities with different implementations? Decide which wins.
4. **Plan file layout** — map each source's files into the target's directory structure

**Common fusion pattern**: Target identity + strongest workflow as backbone + specialized capabilities from others.

### Phase 3 · Execute File Merge

Copy files from all sources into the target skill directory:

```bash
# References (merge, don't overwrite existing)
cp -rn /source/references/*.md target/references/

# Scripts
mkdir -p target/scripts && cp -n /source/scripts/* target/scripts/

# Assets (templates, fonts, etc.)
mkdir -p target/assets && cp -rn /source/assets/* target/assets/
```

**Rules**:
- `-n` (no-clobber) for references — don't overwrite target's existing files
- `-rn` for assets — merge directories but don't overwrite existing files
- Preserve directory structure from source when it adds organizational value
- Check disk space before copying (large skills can be 40MB+)

### Phase 4 · Rewrite SKILL.md

The fused SKILL.md must:

1. **Preserve target's identity** — the name, core philosophy, and design principles stay
2. **Add a capability overview table** — show what came from where (traceability)
3. **Restructure workflows** — integrate absorbed capabilities as new paths/modes
4. **Update the resource map** — the file tree at the bottom must reflect all new files
5. **Keep it under 500 lines** — use references/ for deep detail, SKILL.md for navigation

**Writing the capability table** (critical for traceability):

```markdown
## ⚡ Capability Overview

| Capability | Source | Key Advantage |
|------------|--------|---------------|
| Original feature | target-skill | Already existed |
| New feature A | source-a | 40 style library |
| New feature B | source-b | Quality checklist |
```

**Structural patterns for fused skills**:
- **Task routing table** at the top: "If input is X → follow Y path"
- **Shared principles first**, then path-specific workflows
- **Resource file tree** at the bottom as a map

### Phase 5 · Verify Completeness

After fusion, run a completeness check:

```python
# Extract all file references from SKILL.md
# Check each one exists in the skill directory
# Report missing files as blockers
```

**Verification checklist**:
- [ ] All files referenced in SKILL.md exist on disk
- [ ] All referenced scripts are executable (or have clear invocation instructions)
- [ ] Templates are complete (not truncated during copy)
- [ ] No duplicate files with different content
- [ ] Directory structure is clean (no `.git/`, `node_modules/`, temp files)
- [ ] Total size is reasonable (flag if >100MB)

### Phase 6 · Register and Test

1. Verify the skill loads via `skill_view(name=<target>)`
2. Check that `linked_files` correctly lists all references/assets/scripts
3. Confirm `readiness_status: available`

---

## Ecosystem-Level Competitive Analysis

When the user asks to "compare against competitors" or "对标竞品" across an entire ecosystem (multiple repos + skills), use this workflow **before** the fusion phases above.

From 503496348-ops/mischief-production, MIT.