Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt

Use this file to discover all available pages before exploring further.

Claude Code is selftune’s primary platform with full hook-based integration.

​
Prerequisites

​
Installation

npx skills add selftune-dev/selftune
Then tell your agent: initialize selftune

​
Hook integration

selftune installs 6 hooks into Claude Code’s settings.json:
HookEventPurpose
prompt-logUserPromptSubmitLogs every user query and detects improvement signals via regex
auto-activateUserPromptSubmitEvaluates activation rules (e.g., too many unmatched queries)
skill-change-guardPreToolUse (Write/Edit)Prevents unreviewed SKILL.md changes
evolution-guardPreToolUse (Write/Edit)Blocks conflicting edits during active evolutions
skill-evalPostToolUse (Read)Tracks skill triggers and classifies invocation type
session-stopStopCaptures end-of-session telemetry, spawns signal-reactive orchestrate

​
Data flow

User query → Claude Code session
  ↓
prompt-log hook captures query (< 5ms)
  ↓
skill-eval hook tracks which skills fire
  ↓
session-stop hook writes session summary
  ↓
selftune sync normalizes into SQLite

​
Signal detection

The prompt-log hook detects user corrections in real-time using regex patterns like:
  • “why didn’t you use X?”
  • “you should have used the Y skill”
  • “use Z next time”
When detected, the mentioned skill gets a priority boost for the next orchestrate run.

​
Session paths

PathPurpose
~/.claude/Claude Code session logs
~/.claude/settings.jsonHook configuration
~/.selftune/selftune.dbNormalized selftune data

​
Ingest command

selftune ingest claude [--since DATE] [--dry-run] [--force] [--verbose]
See CLI reference for the full command list.