agentleFS
Sign inStart free
IamLumiRayIamLumiRay/Agent-delegate/references/claude.md

Agent-delegate / references

CLAUDE.md4 starsChanged 15 days agoMITView on GitHub

Use the installed Claude CLI with its existing login. No Claude desktop window, API-key extraction or Python SDK is needed. Register the selected CLI and configuration directory with setup.md. Authentication stays native; the helper checks the CLI's available nonsecret auth metadata before dispatch and on continuation. The fingerprint does not hash credentials and cannot detect changes absent from that metadata. The CLI must support --safe-mode, --restricted, --permission-prompts and structured print output; doctor checks these features. On Windows the official npm…

  • In this repositoryOne other file it gives its agents.

The file

We split this claude.md into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Default tools: Read, Glob and Grep.

    line 27

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
claude.md

Claude Code delegation

Runtime and authentication

Use the installed Claude CLI with its existing login. No Claude desktop window, API-key extraction or Python SDK is needed. Register the selected CLI and configuration directory with setup.md. Authentication stays native; the helper checks the CLI's available nonsecret auth metadata before dispatch and on continuation. The fingerprint does not hash credentials and cannot detect changes absent from that metadata.

The CLI must support --safe-mode, --restricted, --permission-prompts and structured print output; doctor checks these features. On Windows the official npm wrapper resolves to its native claude.exe. Other Windows command wrappers are not executed through a shell.

Run and continue

Configure state ignore rules first. Resolve <skill> to its installed directory.

python -B "<skill>/scripts/invoke_claude.py" new --target claude-main --cwd "<workspace>" --state-dir "<workspace>/.agent-delegate/claude" --prompt "Review these files without editing them."
python -B "<skill>/scripts/invoke_claude.py" send --target claude-main --conversation-id "<UUID>" --state-dir "<workspace>/.agent-delegate/claude" --prompt "Verify the specific concern in your previous answer."
python -B "<skill>/scripts/invoke_claude.py" result --conversation-id "<UUID>" --state-dir "<workspace>/.agent-delegate/claude"

The shared PowerShell entry uses -Provider claude -Target claude-main, with -Workspace, -StateDir, -PromptFile, -PythonPath and -TimeoutSeconds. -ClaudeCli and -ClaudeHome select explicit routing during setup. --model selects a model for a new task; continuation uses the recorded exact resolved model. Optional --effort retains the CLI's supported effort values.

Receipts retain final visible text and routing/status metadata. Thinking, tool payloads, raw errors and transcripts are not copied. Success requires the expected session/workspace, a confirmed model, a successful native result, a zero exit code, no permission denials and a nonempty answer. Verify the actual deliverable.

If the CLI needs proxy or TLS/runtime environment settings, the MCP client must retain the necessary local settings for its server process. An SDK test client may omit them even when native login checks pass. Keep those values local; never put them in the release or target registry.

Tools and lifecycle

  • Default tools: Read, Glob and Grep. Authorized --sandbox workspace-write adds Edit and Write. Native restricted mode confines built-in file tools to working directories; this is not an additional OS sandbox.
  • Safe mode disables customizations while preserving authentication. Restricted mode, explicit tool lists and empty strict MCP configuration exclude shell execution, subagents and external MCP tools. Native administrator policy still applies. The helper does not bypass permissions or fabricate approval.
  • Ordinary tasks and exact UUID continuation are supported. Other providers' modes and workflows are rejected. Never use latest-session selection, name matching, forks or unrelated transcripts as a substitute for the saved ID.
  • Each call owns one print process. Timeout/cancellation stops that process; without a native completion acknowledgement, its receipt is unknown. Inspect the provider before retrying overlapping work. Only completed receipts can be continued by this helper. Locks prevent concurrent helper turns.

References: CLI reference, programmatic use.

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.

Check your setup, free