The Agent Skills Directory

[COMMAND_EXECUTION]: Executes standard diagnostic and maintenance commands using the memsearch CLI, such as config list, stats, and version. It also uses common utilities like uv, git, and ls to inspect the local environment and memory directories.
[EXTERNAL_DOWNLOADS]: Fetches package metadata and documentation from official sources, including PyPI, the vendor's GitHub repository (zilliztech/memsearch), and official documentation domains. These downloads are used for version verification and update guidance.
[REMOTE_CODE_EXECUTION]: Contains a command pattern that downloads JSON metadata from PyPI and pipes it to a Python one-liner (curl ... | python3 -c ...) to extract the package version. This is a targeted data extraction from a well-known service rather than arbitrary code execution.
[INDIRECT_PROMPT_INJECTION]: Implements defensive configuration handling by explicitly allowlisting only low-risk indexing keys for project-local configuration files. Sensitive settings like provider credentials and prompt paths are restricted to global configuration to prevent repository-based configuration hijacking.
[CREDENTIALS_SAFE]: Promotes secure secret management by instructing users to use environment variable references (e.g., env:OPENAI_API_KEY) for provider authentication instead of hardcoding sensitive tokens.

memory-config