Claude-skill-registry alto-configure

Use when configuring ALTO settings including arbiter thresholds, permissions, or verification hooks. Applies during setup orchestrator "Configure ALTO" selection, build orchestrator checkpoints, or when user requests configuration changes.

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/alto-configure" ~/.claude/skills/majiayu000-claude-skill-registry-alto-configure && rm -rf "$T"
manifest: skills/data/alto-configure/SKILL.md
source content

ALTO Configuration

Shared configuration procedures for setup and build orchestrators.

Arbiter Thresholds

Use

AskUserQuestion
:

  • Header: "Autonomy"
  • Question: "How much autonomy should ALTO have before checkpoints?"
  • Options:
    1. Label: "Conservative", Description: "500 lines, 20 files, checkpoint every 2 tasks"
    2. Label: "Balanced (Recommended)", Description: "2000 lines, 50 files, checkpoint every 3 tasks"
    3. Label: "Autonomous", Description: "5000 lines, 100 files, checkpoint every 5 tasks"

After user answers, read

alto.json
, update the
arbiter
section, and write back:

Selectionmax_linesmax_filestask_interval
Conservative500202
Balanced2000503
Autonomous50001005

Procedure:

  1. Read
    alto.json
  2. Update
    arbiter.max_lines_changed_without_human
    ,
    arbiter.max_files_changed_without_human
    ,
    arbiter.task_checkpoint_interval
  3. Write back
    alto.json

Example updated

arbiter
section for "Balanced":

"arbiter": {
  "max_lines_changed_without_human": 2000,
  "max_files_changed_without_human": 50,
  "task_checkpoint_interval": 3,
  ...
}

Permissions

Use

AskUserQuestion
:

  • Header: "Permissions"
  • Question: "What permission level for bash commands?"
  • Options:
    1. Label: "Supervised (Recommended)", Description: "Prompt for git, npm, docker"
    2. Label: "Autonomous", Description: "Auto-approve most commands"
    3. Label: "Locked", Description: "Prompt for everything"

If NOT "Supervised": Tell user to update

devenv.nix
:

alto.permissions.profile = "autonomous";  # or "locked"

Then

alto-restart
. (Permissions require Nix change - orchestrator cannot modify at runtime.)

Verification (Existing Projects Only)

Read

alto.json
, display the
verification
section, then use
AskUserQuestion
:

  • Header: "Verification"
  • Question: "How would you like to adjust verification?"
  • Options:
    1. Label: "Keep current", Description: "No changes"
    2. Label: "Add pattern", Description: "Add verification for new file type"
    3. Label: "Edit command", Description: "Change existing command"
    4. Label: "Remove pattern", Description: "Remove file type verification"

Procedure: Read

alto.json
, update the
verification
section, write back
alto.json
.

Format for verification section:

"verification": {
  "<glob-pattern>": {
    "<check-type>": "<command>"
  }
}

Example:

"verification": {
  "*.ts": { "typecheck": "npx tsc --noEmit", "lint": "npm run lint" },
  "*.py": { "lint": "ruff check {file}" },
  "*.test.ts": { "test": "npm test -- --related" }
}

Check types:

lint
,
typecheck
,
test
,
format

New Projects

Skip verification. Say: "Verification starts empty. QA agent configures it when tooling is set up."

Key Rules

  1. Orchestrator writes JSON files - never ask user to edit JSON manually
  2. Orchestrator confirms changes after writing
  3. Only permissions require user edit (devenv.nix + alto-restart)