Skip to content
Clasper Core

Workspace

Workspace

Section titled “Workspace”

Clasper Core uses a workspace folder for portable, version-controlled agent configuration.

Workspaces define:

  • agent personas and operating rules
  • optional multi-agent role personas
  • optional skills (instructions + capability intent)

Workspace location

Section titled “Workspace location”

Configure via:

Terminal window
CLASPER_WORKSPACE=./workspace

File structure

Section titled “File structure”
workspace/
├── AGENTS.md            # Operating rules (recommended)
├── SOUL.md              # Default persona (fallback)
├── souls/               # Role personas (optional)
│   └── <role>.md
├── IDENTITY.md          # Branding (optional)
├── HEARTBEAT.md         # Heartbeat checklist (optional)
└── skills/              # Skills (optional)
    └── <skill-name>/
        └── SKILL.md

Key files

Section titled “Key files”

AGENTS.md

Section titled “AGENTS.md”

Shared operating rules: constraints, safety rules, and execution style. Treat this like policy and review changes like code.

SOUL.md and souls/<role>.md

Section titled “SOUL.md and souls/<role>.md”

Persona definitions. If souls/<role>.md exists, it is used for that role; otherwise SOUL.md is the default.

skills/<name>/SKILL.md

Section titled “skills/<name>/SKILL.md”

Skill instructions (OpenClaw-compatible format). Skills are a convenient way to:

  • package instructions for consistent behavior
  • describe intended capabilities and boundaries

Governance note

Section titled “Governance note”

In v2.0, execution is mediated by governance and (optionally) adapters. Workspace content is one of the inputs into safe, repeatable execution, but it does not grant authority by itself.

Section titled “Related”