Where Does Distilly Store Local Data? Directory Structure Explained

Distilly stores configuration files in ~/.distilly/ and skill metadata in hidden .distilly-install.json files within each skill's installation directory.

Distilly, an open-source project under the titanwings organization, persists data across two distinct locations on a user's machine. Understanding where Distilly stores local data is essential for troubleshooting, backups, and manual configuration of its chat platform collectors. The project separates global configuration from skill-specific metadata using a predictable directory structure.

Configuration Storage: The ~/.distilly/ Directory

All collector tools reference a constant called CONFIG_PATH that resolves to the .distilly folder in the user's home directory. As implemented in tools/slack_auto_collector.py, tools/feishu_auto_collector.py, and tools/dingtalk_auto_collector.py, this path stores JSON configuration files for each respective platform.

The typical configuration files include:

• slack_config.json – API tokens and workspace IDs for Slack
• feishu_config.json – Feishu platform credentials
• dingtalk_config.json – DingTalk integration settings
• config.json – Global CLI configuration and migration state

The directory is excluded from version control via the .gitignore entry /.distilly/, ensuring sensitive tokens are never committed.

Loading Collector Configuration

To load a collector's configuration, the tools use standard path resolution:

from pathlib import Path
import json

# Example: load Slack config

CONFIG_PATH = Path.home() / ".distilly" / "slack_config.json"

def load_slack_config() -> dict:
    if CONFIG_PATH.is_file():
        return json.loads(CONFIG_PATH.read_text())
    raise FileNotFoundError(f"Slack config not found at {CONFIG_PATH}")

slack_cfg = load_slack_config()
print(slack_cfg["bot_token"])

Source: tools/slack_auto_collector.py

Skill Installation Metadata

When Distilly installs a skill, it writes a hidden metadata file alongside the skill definition. According to tools/install_generated_skill_common.py, the installer creates .distilly-install.json inside the skill directory (e.g., ~/.agents/skills/distilly/.distilly-install.json).

This JSON file records:

• preset: The generation preset used (e.g., "distilly.colleague.v1")
• engine: Object containing {"name": "distilly"}
• generation: Object tracking the generation engine

The installer also uses a temporary staging folder named .distilly-install-<pid> during the copy process, but only the final .distilly-install.json persists.

Reading Skill Metadata

The metadata file enables downstream tools to verify how a skill was generated:

from pathlib import Path
import json

def read_install_metadata(skill_dir: Path) -> dict:
    meta_path = skill_dir / ".distilly-install.json"
    return json.loads(meta_path.read_text())

# Suppose the skill was installed to ~/.agents/skills/distilly

skill_path = Path.home() / ".agents" / "skills" / "distilly"
metadata = read_install_metadata(skill_path)
print(metadata["preset"])               # e.g. "distilly.colleague.v1"

print(metadata["engine"]["name"])       # always "distilly"

Source: tools/install_generated_skill_common.py

Global Configuration Migration

During CLI upgrades, Distilly may create a top-level config.json inside ~/.distilly/ to track migration state. As shown in tests/test_config_migration.py, this file contains a source field marking the origin of installed skills:

{"source": "distilly"}

This allows the CLI to distinguish Distilly-generated skills from other sources.

Enumerating Installed Skills

Since skills can reside in multiple locations (e.g., ~/.agents/skills/, ~/.openclaw/workspace/skills/, ~/.hermes/skills/openclaw-imports/), tools scan for the .distilly-install.json marker to identify Distilly-managed installations:

from pathlib import Path
import json

def list_distilly_skills(base_dirs):
    """Yield (skill_path, metadata) for each installed Distilly skill."""
    for base in base_dirs:
        for skill_dir in Path(base).glob("*"):
            meta_file = skill_dir / ".distilly-install.json"
            if meta_file.is_file():
                yield skill_dir, json.loads(meta_file.read_text())

bases = [
    Path.home() / ".agents" / "skills",
    Path.home() / ".openclaw" / "workspace" / "skills",
    Path.home() / ".hermes" / "skills" / "openclaw-imports",
]

for path, meta in list_distilly_skills(bases):
    print(f"{path} – preset: {meta['preset']}")

Source: Logic derived from tests/test_install_openclaw_and_codex.py and related installer test suites.

Summary

• Distilly stores global configuration in ~/.distilly/ (e.g., API keys for Slack, Feishu, and DingTalk collectors).
• Each installed skill contains a hidden metadata file (.distilly-install.json) recording the generation preset and engine details.
• The ~/.distilly/ directory is excluded from Git via .gitignore to protect sensitive credentials.
• Skills may be installed to various agent directories, but all are identified by the presence of .distilly-install.json.

Frequently Asked Questions

Where is the Distilly configuration directory located?

Distilly creates a hidden directory at ~/.distilly/ in the user's home folder. This location is defined by the CONFIG_PATH constant across collector tools in tools/slack_auto_collector.py, tools/feishu_auto_collector.py, and tools/dingtalk_auto_collector.py.

What files are created inside the skill installation directory?

Each skill directory contains SKILL.md (the skill definition) and a hidden .distilly-install.json file. The metadata file stores the preset used, engine name, and generation settings, as written by tools/install_generated_skill_common.py.

How does Distilly handle configuration for different chat platforms?

Each platform uses a separate JSON file within ~/.distilly/: slack_config.json, feishu_config.json, and dingtalk_config.json. These store platform-specific credentials like bot tokens and workspace IDs required by their respective auto-collectors.

Is the .distilly directory tracked in version control?

No. The repository's .gitignore explicitly excludes /.distilly/ to prevent accidental commits of sensitive configuration files containing API tokens and workspace credentials.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}