Quickstart: Install the attune-ai plugin for Claude Code

Add the plugin to Claude Code and confirm its hooks are working.

claude plugin marketplace add Smart-AI-Memory/attune-ai && claude plugin install attune-ai@attune-ai

Result: 13 skills available via slash commands in Claude Code.

Step 1: Confirm the welcome hook fires

Open a new Claude Code session. The hooks.welcome entry point runs automatically on session start. You should see an orientation message in the session pane.

Step 2: Check context utilization

In a session with an existing transcript, the compact-warning hook estimates how full the context window is:

from hooks._transcript_size import estimate_utilization

util = estimate_utilization("~/.claude/transcript.json")
print(util)  # e.g. 0.73

Result: A float between 0.0 and 1.0. A value above your configured threshold triggers the warning produced by hooks.compact_warning.

Step 3: Orient to in-flight specs

Run the spec-orient hook to see which specs are active in your workspace:

from hooks._state import workspace_roots, discover_specs
from pathlib import Path

roots = workspace_roots(Path.cwd())
specs = discover_specs(roots)
for s in specs:
    print(s.slug, s.phase, s.status)

Result: One line per in-flight spec, for example:

my-feature draft in-progress

No output means no active specs were found under the workspace roots.

Step 4: Verify the security guard

Confirm the security guard rejects a disallowed path before you rely on it in automation:

from hooks.security_guard import validate_file_path

ok, reason = validate_file_path("/etc/passwd")
print(ok, reason)  # False  <rejection reason>

Result: ok is False and reason explains the rejection. A path outside SYSTEM_DIRECTORIES that you own returns True.

Next: Type /attune in a Claude Code session to run your first skill.

Unresolved references

Auto-generated by attune-author fact-check. Review and either fix the source code, fix this doc, or add an override.