Wrapper Skill Design
Input: $ARGUMENTS
Step 1: Define What Gets Wrapped
A wrapper skill takes another skill’s output and makes it better.
WRAPPER NAME: [abbreviation]
WRAPS: [what kind of skill output it enhances]
ADDS: [what new layer it provides]
PRESERVES: [what it keeps unchanged from the original]| Wrapper Type | What it adds | Example |
|---|---|---|
| Validator | Checks the output for errors | ”Verify all claims have evidence” |
| Enricher | Adds missing information | ”Add confidence scores to each item” |
| Challenger | Stress-tests the output | ”Find the weakest point in the analysis” |
| Formatter | Restructures for a new audience | ”Turn analysis into executive summary” |
| Synthesizer | Combines outputs from multiple skills | ”Merge perspectives into unified view” |
Step 2: Define the Wrapping Contract
Specify what goes in and what comes out.
WRAPPING CONTRACT:
- Accepts: [types of skill output this wrapper can handle]
- Minimum input: [what must be present for the wrapper to work]
- Rejects: [what it can't wrap — be explicit]
OUTPUT:
- Original output: [preserved in full / summarized / referenced]
- Added layer: [what the wrapper contributes]
- Combined format: [how original + new layer are presented]Rule: The original output must always be recoverable. A wrapper adds — it never destroys.
Step 3: Design the Enhancement Layer
Define what the wrapper does to the wrapped output.
ENHANCEMENT STEPS:
1. RECEIVE: [accept the wrapped skill's output]
2. PARSE: [identify the key elements to enhance]
3. ANALYZE: [apply the wrapper's specific lens]
- [sub-analysis 1]
- [sub-analysis 2]
- [sub-analysis 3]
4. ANNOTATE: [attach findings to the original output]
5. DELIVER: [present original + annotations together]Constraint: Steps 2-4 should be completable without re-running the wrapped skill.
Step 4: Handle Multiple Wrapped Skills
A good wrapper works with many different skill outputs.
COMPATIBILITY MATRIX:
| Wrapped Skill | What Wrapper Adds | Works Well? |
|--------------|-------------------|-------------|
| [skill 1] | [specific enhancement] | [yes/partial/no] |
| [skill 2] | [specific enhancement] | [yes/partial/no] |
| [skill 3] | [specific enhancement] | [yes/partial/no] |
| [skill 4] | [specific enhancement] | [yes/partial/no] |If compatibility is “partial,” note what’s missing. If compatibility is “no,” the wrapper scope may be too narrow.
Step 5: Preserve and Append
Design the output format that shows original and enhancement clearly.
OUTPUT FORMAT:
## Original Output
[The wrapped skill's complete output, unchanged]
## [Wrapper Name] Enhancement
[The wrapper's added analysis/validation/enrichment]
### Key Findings
- [finding 1]
- [finding 2]
- [finding 3]
### Recommendations
- [action 1 based on wrapper analysis]
- [action 2 based on wrapper analysis]Rule: A reader should be able to ignore the enhancement and still use the original output.
Step 6: Test With Multiple Skills
Verify the wrapper adds value across different skill outputs.
WRAP TEST 1:
- Wrapped skill: [name]
- Original output quality: [assessment]
- After wrapping: [what improved]
- Value added: [concrete benefit]
WRAP TEST 2:
- Wrapped skill: [name]
- Original output quality: [assessment]
- After wrapping: [what improved]
- Value added: [concrete benefit]
WRAP TEST 3:
- Wrapped skill: [name]
- Original output quality: [assessment]
- After wrapping: [what improved]
- Value added: [concrete benefit]Integration
Use with:
/chns-> Design the wrapper as a chainable post-processing step/injc-> Consider if the wrapper could also work as a lightweight injectable/stnl-> Ensure the wrapper is also useful when given raw input (not just skill output)