examples
How to Install
Claude Code:
git clone --depth 1 https://github.com/emmahyde/dotfiles.git && cp dotfiles/.claude/skills/source-to-skill/examples ~/.claude/skills/examples -r# BAD: jq CLI → common failure modes
Same source as good-jq-skill.md, showing what goes wrong.
---
## Failure 1 — Vague triggers (most common failure)
```markdown
description: "Helps with JSON processing and data manipulation tasks. Use when working with JSON data or command line tools."
```
Why it fails: "JSON processing", "data manipulation", "command line tools" — Claude will trigger this for ANY JSON task, including Python scripts, curl calls, and database queries. The source's own terms (`jq`, `.[]`, `select`, `map`) appear nowhere. Claude can't distinguish this from a generic JSON skill.
Fix: derive triggers from the source's own command names and syntax symbols.
---
## Failure 2 — Wrong freedom level (high for a low-freedom tool)
```markdown
## Working with arrays
When you need to process arrays in jq, you can use various array-processing techniques. Consider the structure of your data and choose the appropriate filter based on what you're trying to accomplish. The select() function is useful for filtering, while map() helps with transformations.
```
Why it fails: jq has exact, non-negotiable syntax. "Various techniques" and "choose appropriate filter" gives Claude no actionable commands. A user asking "how do I filter this array" gets a paragraph, not `jq '.[] | select(.active == true)'`. Low-freedom tools need literal commands.
Fix: every workflow step is a runnable command with the exact syntax.
---
## Failure 3 — Padded workflows (invented, not source-derived)
```markdown
## Workflows
### General JSON workflow
1. Identify your JSON structure
2. Plan your transformation
3. Write your jq filter
4. Test and iterate
### Advanced JSON workflow
1. Consider performance implications
2. Use streaming for large files
3. Combine with other tools
```
Why it fails: "Identify your JSON structure" and "plan your transformation" are not jq — they're generic advice. The source defines three real workflows (extract, filter, transform). These invented steps don't appear in the source and add noise without value.
Fix: only write workflows that exist in the source. If the source has 3, write 3.
---
## Failure 4 — Missing error patterns
A skill with no errors section forces Claude to hallucinate error behavior. jq returns `null` for missing keys (not an error), exits non-zero for parse failures (not a warning), and produces cryptic type messages (`Cannot index string with string "x"`). These are high-frequency support questions that belong in the skill.
Fix: grep the source and GitHub issues for recurring error strings and add them verbatim.
---
## Failure 5 — Time-sensitive content
```markdown
## Installation
Install jq 1.7.1 (latest as of March 2025): `brew install jq`
```
Why it fails: version numbers and "latest" dates rot immediately. The skill becomes wrong without any warning.
Fix: never include version numbers or dates. Just `brew install jq`.
Details
| Category | Data → data_proc |
| Source | emmahyde/dotfiles |
| SKILL.md | View on GitHub → |
| Repo Stars | N/A |
| Est. per Skill | N/A (shared across 91 skills from this repo) |
| Difficulty | Intermediate |
| Risk Level | Safe |
Related Skills
monte-carlo-prevent
Monte Carlo Prevent Skill This skill brings Monte Carlo's data observability context directly into y
sql-pro
You are an expert SQL specialist mastering modern database systems, performance optimization, and ad
fp-data-transforms
Practical Data Transformations This skill covers the data transformations you do every day: working
options-flow-analyzer
Options Flow Analyzer Analyze options chain data with real vs lottery call separation — the key insi
Works Well With
Skills from the same repository — often designed to work together
duet
--- name: duet description: Two-party posture — user as director, agent as executor; every fork, tra
audit
--- description: >- Structured PASS/FAIL audit of any target against any criteria. Takes a subject a
buddy
--- name: buddy description: "Show, pet, or manage your coding companion. Use when the user types /b