Please adhere to a better skills standard
- Dominant language
- TypeScript
- Stars
- 631
- Forks
- 100
- Avg merge
- 5h 9m
- Merged PRs (30d)
- 25
Description
firecrawl-cli is a single tool and therefore it should NOT require 33 skills.
This adds a huge amount of redundant crap into the context window of every user submitted message. Additionally it's incredibly annoying and messy having a tool that just spams files and folders all over the ~ directory for every harness you support.
Using a tool should be only a single skill and you should follow a better standard.
A SKILL.md should represent the 4 W's:
- What
- Where
- When
- Why
The references should contain the how
Please create a single skills (firecrawl-cli) and structure it like this:
```md
---
name: firecrawl-cli
description: ...
allowed-tools:
- Bash(firecrawl *)
- Bash(npx firecrawl-cli *)
metadata:
- last-updated: X
---
TheFirecrawl CLI enables easy interaction with the web.
- [firecrawl search {query}](./references/firecrawl-search.md)
- Use When:
- Url Unknown, finding sources, articles or news.
- Not When:
- Research papers: `firecrawl research index *` better
- Software development (libs, api, error, bugs, etc): `firecrawl developer index *` better
- ...etc
```
Better yet, create a simple recommender system that the agent can use via the cli. Like `firecrawl skill recommender *query` for eg: `firecrawl skill recommender "firecrawl api search executeSearch command"` would return `[firecrawl developer](./references/firecrawl-developer-index.md)` and `[firecrawl search](./references/firecrawl-search.md)`
Either way, please stop creating N^10 skills for a single product.
Useful refs for you to check out:
- https://agentskills.io/home
Thanks
Contributor guide
No contributing guide indexed for this repository
Research direction
The issue names no repository files or tests; start by inventorying the current 33 skills and the existing firecrawl-cli skill layout, then compare the proposed SKILL.md and references structure with agentskills.io. Done means redundant per-harness skills are consolidated, references cover the documented commands, and any recommender behavior is specified and tested.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- markdown
- Domain
- cli, documentation, tooling
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100