settings
Skill1 starsChanged 5 months agoMITView on GitHub
Configure, audit, or maintain settings.json and settings.local.json. Use when adding permissions, model selection, env vars, output style, or any other Claude Code runtime configuration.
- Reads credentials
9 other projects use a version of this.
Adopted by
Most recently taken up by
- buildomator/buildomatortheir own version, first seen Nov 4, 2026
- tjmode/koovitheir own version, first seen Nov 1, 2026
- aibtcdev/skillstheir own version, first seen Oct 26, 2026
- Donchitos/Claude-Code-Game-Studiostheir own version, first seen Oct 19, 2026
- venus2099/codex-game-studiostheir own version, first seen Oct 17, 2026
- wfukatsu/slide-forgetheir own version, first seen Sep 17, 2026
Used with
Projects with it also use help.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
MicrosoftSkillhelp
Entry point for the MedImageInsight plugin, covering the getting-started walkthrough, how the skills work, what MI2 is, and troubleshooting.
Also in 3 projects that use this
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- Donchitos/Claude-Code-Game-Studios★ 27ksame name
- buildomator/buildomator★ 102same name
- crisandrews/ClawCode★ 64same name
- liamcain/obsidian-dev-skills★ 16same name
- rvdbreemen/adr-kit★ 13same name
- aibtcdev/skills★ 11same name
- wfukatsu/slide-forge★ 5same name
- tjmode/koovi★ 1same name
- venus2099/codex-game-studios★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository26 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- audit-legalSkill
- audit-securitySkill
- echartsSkill
- explainSkill
- file-changelogSkill
- file-claudeSkill
- file-licenseSkill
- file-projectSkill
- file-readmeSkill
- frontendSkill
- hooksSkill
- lean-reviewSkill
- loopSkill
- masterpromptSkill
- mcpSkill
- mermaidSkill
- new-projectSkill
- outputstyleSkill
- proofreadSkill
- rulesSkill
- skillsSkill
- sotaSkill
- subagentsSkill
- tasksSkill
- toonSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 5 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 22
Never put behavioral instructions, style guides, or workflow guidance in settings.json — those belong in CLAUDE.md, rules, or skills.
- Very specific: line 24
Never commit secrets, tokens, or machine-specific absolute paths to shared project settings (.claude/settings.json).
- Very specific: line 26
Never add credential paths (~/.aws/credentials, .env, secrets/) to allow without explicit user acceptance of the security risk.
- Distinctive: line 28
Never produce partial JSON — always deliver a complete, valid file.
- Very specific: line 30
Always include $schema in produced files.
- Very specific: line 31
Never author hook script logic here — the settings.json hooks field only wires events to companion scripts.
- Distinctive: line 37
Always-relevant project context (commands, architecture, gotchas): that is always-loaded context (CLAUDE.md).
- Distinctive: line 92
Read the target file if it exists.
- Distinctive: line 95
Write the complete updated file.
- Very specific: line 99
Read all accessible settings files: settings.local.json, .claude/settings.json, ~/.claude/settings.json.
- Distinctive: line 108
Read the target file.
“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Settings JSON
1§ Purpose
Produce valid, complete settings.json or settings.local.json files for Claude Code. Covers permissions, model selection, environment variables, and all other runtime configuration fields. Not for behavioral guidance — that belongs in rules or CLAUDE.md.
2§ Use when
- User wants to add, change, or remove a setting (permissions, model, env, etc.).
- User wants to audit an existing settings file for stale rules, conflicts, or risks.
- User wants to clean up accumulated one-off permission entries.
- User says "configure settings", "add permission", "allow/deny tool", "set model", "settings.json", or "settings.local.json".
3§ Hard constraints
- Never put behavioral instructions, style guides, or workflow guidance in settings.json — those belong in CLAUDE.md, rules, or skills.
- Never commit secrets, tokens, or machine-specific absolute paths to shared project
settings (
.claude/settings.json). - Never add credential paths (
~/.aws/credentials,.env,secrets/**) toallowwithout explicit user acceptance of the security risk. - Never produce partial JSON — always deliver a complete, valid file.
- Preserve all existing entries not in conflict with the request.
- Always include
$schemain produced files. - Never author hook script logic here — the settings.json
hooksfield only wires events to companion scripts.
4§ Decision gate
- Authoring hook scripts (the logic behind a
hooksentry): out of scope — that is a separate hook artifact. - MCP server configuration (
command,args,url,envper server): that belongs in.mcp.json, not settings. - Behavioral guidance ("Claude should...", style rules): that is a rule.
- Always-relevant project context (commands, architecture, gotchas): that is always-loaded context (CLAUDE.md).
- Unknown field: check
$schemaathttps://json.schemastore.org/claude-code-settings.json. - Scope unclear: default to
settings.local.json, label[assumed], explain the trade-off.
Scope selection:
| Applies to | File |
| This machine only, not shared | .claude/settings.local.json |
| Team-shared via git | .claude/settings.json |
| Personal defaults across all projects | ~/.claude/settings.json |
| Org-wide enforcement (MDM only) | managed-settings.json |
Precedence (highest → lowest): Managed → CLI args → Local → Project → User.
Arrays (permissions, hooks, availableModels) merge across scopes. Scalars override.
5§ Permission rules
The permissions object has three levels, evaluated in order — first match wins:
deny— blocked entirely, no user prompt possibleask— always prompts, even if previously approvedallow— pre-approved, never prompts
Rule pattern syntax:
| Pattern | Matches |
Bash(cmd *) | Bash commands starting with cmd (* = any args, single level) |
Read(./path/**) | File reads under a path (** = recursive) |
Edit(./path/**) | File edits under a path |
Write | All file writes |
WebFetch(domain:example.com) | Web requests to a specific domain |
WebFetch(domain:*.example.com) | Domain wildcard (all subdomains) |
MCP(server-name) | All tools from a specific MCP server |
mcp__server-name__* | All tools from a specific MCP server (alternative syntax) |
mcp__server-name__tool | One specific MCP tool |
Agent | All subagent spawns |
Agent(type) | Spawning a specific subagent type |
Common deny patterns:
| Goal | Deny rule |
| Block all computer-use (screenshots, clicks, keyboard) | mcp__computer-use__* |
| Block all browser MCP tools | mcp__Claude_in_Chrome__* |
| Block subagent spawning entirely | Agent |
| Block a specific subagent type | Agent(explore) |
Additional permission fields:
additionalDirectories— grants Read to paths outside the project rootdefaultMode: "acceptEdits"— pre-approves all Write/Edit without per-call promptsdisableBypassPermissionsMode: "disable"— blocks--dangerously-skip-permissions
6§ Workflow
6.1§ Configure (add or change a field)
- Read the target file if it exists; note if a new file will be created.
- Apply the decision gate — confirm settings.json is the right artifact.
- Determine scope (§4). Label
[assumed]if defaulting. - Write the complete updated file. Preserve all existing entries not in conflict.
- Confirm: output the file path and a one-line summary of what changed.
6.2§ Audit (review an existing file)
- Read all accessible settings files:
settings.local.json,.claude/settings.json,~/.claude/settings.json. - Flag: overly broad permissions, credential paths in
allow, machine-specific absolute paths in shared files, behavioral prose, hook entries without companion scripts, scope conflicts (same key overriding unintentionally across files). - Report by severity: Blocker (security risk) → Warning (maintenance) → Note.
- Propose fixes; apply only after user confirmation.
6.3§ Maintain (clean up)
- Read the target file.
- Identify stale entries (commands/paths no longer in the project) and one-off permissions that can be consolidated or removed.
- Propose a cleaned version; apply after user confirmation.
7§ Output contract
Must include:
$schemaas the first key in every produced file- Complete file content — never a partial snippet requiring manual merging
- Confirmation: file path and one-line summary of changes
Must not include:
- Comments inside
.jsonfiles (RFC 8259 prohibits them; use.jsoncfor annotated templates only) - Trailing commas
- Behavioral instructions, prose, or style guidance
- Secrets, tokens, or credential values
- Machine-specific absolute paths in shared project files
8§ Associated documents
- [assets/settings.example.json] — example of a well-structured project settings.json.
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.

