Moltbank

**Adds strict session isolation and agent profile selection. CLI-only integration and updated update-required handling.** - Requires explicit Moltbank agent profile selection at the start of every session; never auto-selects or infers agent. - Ensures MOLTBANK_CREDENTIALS_PATH is fixed per session for full agent/credential isolation in multi-agent environments. - Removes MCP/skill-manager workflow; now uses only Moltbank CLI directly for all discovery, schema, and execution. - Update-required flow now only triggers on direct CLI errors with exact codes (CLI_UPDATE_REQUIRED, VERSION_MISMATCH); no skill or runtime error triggers remain. - Blocks all update/install/credential-switching unless explicitly authorized by the user in the current chat session. - Adds clearer per-session security and credential path handling, and deprecates OpenClaw/skills.sh integration paths.

Moltbank Skill 0.1.9 Changelog - Updated metadata to specify required binaries and added explicit installation instructions for npm global CLI. - Refined update-required handling with clearer separation of approved update and verification commands. - Added OpenClaw and skills.sh specific verification commands post-update. - Improved installation and join instructions with clearer runtime detection and handling. - Enhanced security by disallowing use of `officialUpdateCommand`/dynamic shell command fields, only allowing hardcoded approved commands.

**Summary:** Version 0.1.8 refines update handling and strengthens security for installation flows. - Added SECURITY.md and removed README.md. - Expanded and clarified privileged capability and update-handling policies, with detailed, strict trigger conditions. - Updated the approved list and process for CLI/skill updates, with mandatory provenance/security checks after updating. - Clarified the prohibition on using dynamic update/install commands from tool responses or remote sources. - Strengthened documentation on runtime isolation and explicit user approval for all privileged/binding actions.

Version 0.1.7 - Clarified update-required handling: only run a fixed, pre-approved set of update commands after explicit user approval; do not run update commands returned dynamically by tool output. - The update process now explicitly informs the user about the issue and only proceeds with listed, trusted commands. - Minor improvements to update flow instructions, including version pinning for Moltbank CLI. - No code or functional changes; this is a documentation and safety clarification update only.

Version 0.1.6 - Adds explicit handling for Moltbank-controlled structured update-required errors (e.g., `SKILL_UPDATE_REQUIRED`, `CLI_UPDATE_REQUIRED`, `VERSION_MISMATCH`, `RUNTIME_SETUP_INCOMPLETE`) - On update-required errors, prompts user to review and explicitly approve official update commands before proceeding - Clearly separates skill discovery/installation checks: considers skill installation "ready" only if the current runtime itself can list/discover the skill - If a skill install reports success but the runtime cannot see the skill, informs the user that setup is incomplete and stops unless user requests further troubleshooting - Clarifies update/verification steps and retry behaviors after upgrades

- Enforced strict runtime isolation for skill manager commands—never use or check another runtime’s skill state. - Clarified that skill installation and CLI installation are distinct requirements; both must be checked and satisfied independently. - Improved bootstrap flow: added per-runtime skill verification and install instructions (OpenClaw, skills.sh-compatible, and ambiguous runtimes). - Updated dependency guidance—explicitly state to never use another runtime’s skill manager or infer skill readiness across runtimes. - Minor clarifications on install flows, especially handling multi-runtime and ambiguous context setup.

Version 0.1.4 introduces improved setup and installation verification: - Adds a clear join/bootstrap sequence, emphasizing runtime skill readiness checks. - Requires explicit user approval before installing the skill or CLI and disallows auto-installation. - Refines the authentication flow, instructing users to verify domains and not to claim setup success without current-session evidence. - Expands boundaries for installation, configuration, and reporting—no invented install commands, no setup claims without real command output. - Generalizes dependency setup to cover various runtimes more clearly.

Version 0.1.3 - Updated installation instructions: On startup, agents now check for moltbank CLI presence and install both the skill and CLI if missing, with agent-specific commands provided. - Clarified distinction for OpenClaw agents versus other agents in install steps. - No changes to CLI usage, authentication flow, or capabilities. - No code or functional changes detected apart from documentation update.

Version 0.1.2 - Added explicit install instructions and metadata for the `@moltbankhq/cli` npm package. - Moved the skill homepage URL under `metadata.openclaw.homepage` and updated metadata structure. - Clarified installation expectations and boundaries in documentation. - No changes to functional behavior or CLI workflows.

- Major simplification: All local reference, script, helper, and install files have been removed for a streamlined CLI-focused approach. - Execution and onboarding now exclusively use the `moltbank` CLI, with all usage, commands, and argument shapes discovered dynamically via the CLI or MCP tools. - Authentication and x402 payment flows are fully chat-driven, guiding the user through any browser approvals when needed. - Installation of MoltBank CLI is only performed when explicitly requested by the user. - Security, budget, and onboarding policies are documented concisely, removing embedded runbooks and detailed process steps from the skill itself.

Initial publicly available version