folder-finder
Skill0 starsChanged 2 months agoMITView on GitHub
This skill enables the LLM to help users find folders by name in a specified directory by producing a strictly-formatted JSON output that describes the search. The JSON output is then parsed by the folder_finder_execute skill to walk the filesystem, match folder names against a pattern, and return the list of matching folder paths. The default working directory is the current directory (./) unless the user specifies otherwise. folder-finder Invoke this skill whenever the user's request implies finding/searching for folders…
- Reads credentials
In this repository14 other files it gives its agents.
- cmd-runnerSkill
- code-runnerSkill
- curlSkill
- db-accessorSkill
- file-changerSkill
- file-creatorSkill
- file-finderSkill
- file-grepSkill
- file-listSkill
- file-readerSkill
- folder-creatorSkill
- folder-listSkill
- scpSkill
- sshSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 5 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Do NOT use this skill when the user wants to:
- Very specific: line 25
Search inside file contents → use file-grep.
- Very specific: line 26
Create a folder → use folder-creator.
- Very specific: line 28
Read a file → use file-reader.
- Distinctive: line 32
When the skill is triggered, the LLM must:
- Very specific: line 96
All string values must properly escape special characters (", \, newline as \n, tab as \t).
- Very specific: line 97
Do NOT emit multiple folder-finder payloads in a single response.
- Distinctive: line 98
If the user does not specify a pattern, do NOT execute this skill — ask for the pattern first or produce an empty action.
- Distinctive: line 206
Never search paths outside the user's intent (no arbitrary system paths).
- Distinctive: line 210
Always echo the JSON payload so the user can review before execution.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
📁 Folder Finder Skill
Overview
This skill enables the LLM to help users find folders by name in a specified directory by producing a strictly-formatted JSON output that describes the search. The JSON output is then parsed by the folder_finder_execute skill to walk the filesystem, match folder names against a pattern, and return the list of matching folder paths.
The default working directory is the current directory (./) unless the user specifies otherwise.
Skill Name
folder-finder
When to Use This Skill
Invoke this skill whenever the user's request implies finding/searching for folders by name. Look for phrases such as:
- "find folders named ..."
- "search for directories with name ..."
- "list folders matching ..."
- "where is the folder ..."
- "locate ... directory"
- "which folders are called ..."
Do NOT use this skill when the user wants to:
- Search inside file contents → use
file-grep. - Create a folder → use
folder-creator. - Modify a folder → (no direct skill).
- Read a file → use
file-reader.
What the LLM Must Do
When the skill is triggered, the LLM must:
- Extract the following from the user's request (asking clarifying questions only if truly ambiguous):
- pattern (required): The folder name or glob pattern to search for.
- path (optional): The directory to search in. Defaults to
./(current working directory). - recursive (optional):
trueto search subdirectories. Defaults totrue. - isRegex (optional):
trueto interpretpatternas a regular expression. Defaults tofalse(glob matching). - caseSensitive (optional):
truefor case-sensitive matching. Defaults tofalse. - includeGlob (optional): A glob pattern to restrict which folders are considered (e.g.
src*). Defaults to*. - excludeGlob (optional): Glob pattern(s) to skip (comma-separated). Defaults to common noise directories like
.git,node_modules,__pycache__,.venv,dist,build. - maxResults (optional): Maximum number of matching folders to return. Defaults to
100. - showPaths (optional):
trueto include full paths in the output. Defaults totrue. - sequential (optional): presented if the find result is needed as input for a follow-up AI task.
- prompt (optional): When
sequentialis presented, this prompt is used to ask the AI to perform all of follow-up tasks on the find result. - description (optional): One-line summary of the search.
- Produce a response containing a single pure JSON object that STRICTLY follows the schema below. No code fences, no extra prose.
Required JSON Output Schema
The LLM's response MUST be a pure JSON string with no code blocks.
{
"action": "find_folder",
"thinking": "<ai-thinking-trace>",
"sequential": {
"prompt": ""
},
"properties": [
{ "name": "pattern", "value": "<folder-name-or-glob>" },
{ "name": "path", "value": "./" },
{ "name": "recursive", "value": "true" },
{ "name": "isRegex", "value": "false" },
{ "name": "caseSensitive", "value": "false" },
{ "name": "includeGlob", "value": "*" },
{ "name": "excludeGlob", "value": ".git,node_modules,__pycache__,.venv,dist,build" },
{ "name": "maxResults", "value": "100" },
{ "name": "showPaths", "value": "true" }
]
}Field Rules
| Field | Type | Required | Notes |
action | string | Yes | Must be exactly "find_folder". |
pattern | string | Yes | The folder name pattern (glob or regex when isRegex=true). Must be non-empty. |
path | string | No | Directory to search. Defaults to ./. |
recursive | boolean | No | Default true. |
isRegex | boolean | No | Default false. |
caseSensitive | boolean | No | Default false. |
includeGlob | string | No | Glob filter for folder names. Default *. |
excludeGlob | string | No | Comma-separated globs to skip. Default excludes common noise dirs. |
maxResults | string | No | Integer as string. Default 100. |
showPaths | boolean | No | Include full paths in output. Default true. |
sequential | object | No | Whether the find result should feed a follow-up AI task. Default null. |
prompt | string | No | Follow-up prompt for all of following tasks when sequential is presented. |
description | string | No | Human-readable purpose of the search. |
Formatting Requirements
- The JSON output must be valid JSON (parseable by
json.loads). - All string values must properly escape special characters (
",\, newline as\n, tab as\t). - Do NOT emit multiple
folder-finderpayloads in a single response. - If the user does not specify a pattern, do NOT execute this skill — ask for the pattern first or produce an empty action.
Response Pattern for the LLM
A typical assistant reply should look like the following. No other output is needed. Output must be in pure JSON format.
{ "action": "find_folder", "thinking": "User wants to find all folders containing 'test' in the project.", "properties": [ { "name": "pattern", "value": "test" }, { "name": "path", "value": "./" }, { "name": "recursive", "value": "true" }, { "name": "isRegex", "value": "false" }, { "name": "caseSensitive", "value": "false" }, { "name": "includeGlob", "value": "*" }, { "name": "excludeGlob", "value": ".git,node_modules,pycache,.venv,dist,build" }, { "name": "maxResults", "value": "100" }, { "name": "showPaths", "value": "true" }, { "name": "description", "value": "Find test folders" } ] }
Decision Logic
User Prompt
|
+-- Does the request imply finding folder/directories by name?
| |
| +-- YES --> Extract pattern/path/flags --> Emit ONE `folder-finder` JSON payload
| |
| +-- NO --> Do not use this skillExamples
Example 1: Simple glob search in current directory
User: "Find all folders named src."
LLM emits:
{
"action": "find_folder",
"thinking": "User wants to find all folders named src in the current directory recursively.",
"properties": [
{ "name": "pattern", "value": "src" },
{ "name": "path", "value": "./" },
{ "name": "recursive", "value": "true" },
{ "name": "isRegex", "value": "false" },
{ "name": "caseSensitive", "value": "false" },
{ "name": "includeGlob", "value": "*" },
{ "name": "excludeGlob", "value": ".git,node_modules,__pycache__,.venv,dist,build" },
{ "name": "maxResults", "value": "100" },
{ "name": "showPaths", "value": "true" },
{ "name": "description", "value": "Find src folders" }
]
}Example 2: Case-sensitive exact name search
User: "Find the folder exactly named config in the project root."
LLM emits:
{
"action": "find_folder",
"thinking": "Case-sensitive exact folder name search for config in root.",
"properties": [
{ "name": "pattern", "value": "config" },
{ "name": "path", "value": "./" },
{ "name": "recursive", "value": "false" },
{ "name": "isRegex", "value": "false" },
{ "name": "caseSensitive", "value": "true" },
{ "name": "includeGlob", "value": "*" },
{ "name": "excludeGlob", "value": ".git,node_modules,__pycache__,.venv,dist,build" },
{ "name": "maxResults", "value": "10" },
{ "name": "showPaths", "value": "true" },
{ "name": "description", "value": "Locate config folder" }
]
}Example 3: Regex search in a specific folder
User: "In src/, find folders matching the regex ^test_.*."
LLM emits:
{
"action": "find_folder",
"thinking": "Regex folder name search in src/ for test folders.",
"properties": [
{ "name": "pattern", "value": "^test_.*" },
{ "name": "path", "value": "src/" },
{ "name": "recursive", "value": "true" },
{ "name": "isRegex", "value": "true" },
{ "name": "caseSensitive", "value": "false" },
{ "name": "includeGlob", "value": "*" },
{ "name": "excludeGlob", "value": ".git,node_modules,__pycache__,.venv,dist,build" },
{ "name": "maxResults", "value": "100" },
{ "name": "showPaths", "value": "true" },
{ "name": "description", "value": "Find test folders in src/" }
]
}Safety Guidelines
- Never search paths outside the user's intent (no arbitrary system paths).
- Refuse suspicious search roots (e.g.,
C:\Windows\,/etc/,~/.ssh/). - Cap the number of folders returned via
maxResultsto avoid runaway output. - Skip unreadable directories silently — do not crash.
- Always echo the JSON payload so the user can review before execution.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

