agentleFS
fmindfmind/dot/skills/agent-harnesses/references/claude.md

dot / references

CLAUDE.md13 starsChanged yesterdayMITView on GitHub

Claude Code sessions, managed settings, headless flags, hooks, and background tasks.

  • Reads credentials

36 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
claude.md

Claude Code (claude)

Operate Claude Code alongside fmind/dot managed configuration; agent-project owns repo bridges, mcp-setup owns MCP servers, and notifications owns turn attention and preserved hooks.

Managed Setup (fmind/dot)

private_dot_claude/modify_private_settings.json merges ~/.claude/settings.json:

  • Seeded defaults (host-overridable via /model, /effort, or /config): model: "fable", effortLevel: "high" (supports low|medium|high|xhigh|max via --effort) on claude-fable-5-1, claude-fable-5-2, and claude-opus-5-5, and remoteControlAtStartup: true.
  • Managed keys (re-applied on chezmoi apply --force): permissions.defaultMode: "bypassPermissions", fallbackModel: ["sonnet"], editorMode: "vim", theme: "light-ansi", tui: "fullscreen", worktree.baseRef: "head", respectGitignore: false, autoContinueAtUsageLimit: true, agentPushNotifEnabled: true, inputNeededNotifEnabled: true, empty commit/PR attribution, disabled claude.ai connectors/plugins/skills sync, and non-interactive CLI env (GIT_TERMINAL_PROMPT=0, GH_PROMPT_DISABLED=1, CLOUDSDK_CORE_DISABLE_PROMPTS=1), preserving host SessionStart hooks and retiring legacy Notification turn hooks.
  • Global links & trust: ~/.claude/CLAUDE.md → ~/.agents/AGENTS.md, ~/.claude/skills → ~/.agents/skills; dot trust manages workspace trust and permissions.additionalDirectories.

Workflow

  1. Run non-interactive (-p / --print):

    claude -p "Summarize failing tests" --output-format json --no-session-persistence
    claude -p "Extract findings" --output-format json --json-schema '{"type":"object","properties":{"ok":{"type":"boolean"}},"required":["ok"]}'

    Pass --no-session-persistence on disposable automation (CLAUDE_CODE_FORCE_SESSION_PERSISTENCE=1 is set in managed env so sub-sessions archive cleanly), and bound unattended runs with --max-turns <N> or --max-budget-usd <amount>.

  2. Isolate clean-room runs with --bare: skips settings/plugin hooks, LSP, CLAUDE.md auto-discovery, auto-memory, and keychain/OAuth reads (ANTHROPIC_API_KEY, apiKeyHelper, or Vertex/Bedrock credentials only); pass --system-prompt, --settings, --mcp-config, and --add-dir explicitly.
  3. Manage background sessions (--bg): claude --bg "<prompt>" prints a session ID controlled via claude agents (lists background sessions, not custom roles), claude logs <id>, claude attach <id>, claude stop <id>, and claude rm <id>.

Gotchas

  • Project skills need .claude/skills: Claude Code ignores .agents/skills/ directly unless .claude/skills -> ../.agents/skills is linked (agent-project).
  • MCP scope privacy: claude mcp add --scope local (default) keeps credentials per-project in ~/.claude.json; --scope project writes shared .mcp.json (requires approval; claude mcp reset-project-choices resets).
  • Never export DISABLE_TELEMETRY=1 globally: Claude Code disables Remote Control when DISABLE_TELEMETRY is set in the environment; prefix DISABLE_TELEMETRY=1 per command only for skills.
  • Gate /ultrareview: invoke /ultrareview only on explicit request due to heavy multi-pass quota cost.

Documentation

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.