The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill requires a wallet private key (MEMOCLAW_PRIVATE_KEY) for authentication and micropayments. It stores this sensitive credential locally in ~/.memoclaw/config.json after running the initialization command.
[EXTERNAL_DOWNLOADS]: Recommends global installation of the memoclaw package from NPM. The CLI includes an upgrade command that allows it to check for and install updates directly from the vendor's infrastructure.
[COMMAND_EXECUTION]: The skill relies extensively on the exec tool to run CLI commands for storing, recalling, and managing memories. It also provides commands to modify shell profile files (~/.bashrc, ~/.zshrc) to enable command-line completions.
[DATA_EXFILTRATION]: The core functionality involves transmitting user-provided text, conversation history, and local file contents (via the --file and migrate commands) to the remote API at api.memoclaw.com. The skill includes clear warnings to avoid storing secrets in this service.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting untrusted data from an external source.
Ingestion points: Untrusted data enters the agent context through the memoclaw recall, memoclaw core, and memoclaw context commands in SKILL.md.
Boundary markers: No specific boundary markers or "ignore instructions" warnings are defined for the retrieved memory content.
Capability inventory: The skill has exec capabilities to run shell commands and access the filesystem via the memoclaw CLI.
Sanitization: No sanitization or validation of the retrieved memory content is described before it is interpolated into the agent's prompt.

memoclaw