The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands including git, uv, pip, and npx to verify environment state, install the core basic-memory package, and install additional skills.
[EXTERNAL_DOWNLOADS]: Fetches the basic-memory tool from standard package registries and downloads supplementary skill definitions from the vendor's repository using npx skills add.
[REMOTE_CODE_EXECUTION]: The installation of external skills from a remote repository via npx skills add constitutes a remote code execution vector used for plugin extension.
[DATA_EXFILTRATION]: Accesses local Git metadata (remotes, branches) and project configuration files (.claude/settings.json) to establish project identity and apply plugin settings.
[PROMPT_INJECTION]: The skill ingests data by reading existing project notes and schemas to infer placement conventions, creating an indirect prompt injection surface where malicious content within notes could influence the setup.
Ingestion points: list_directory, read_note, and schema_infer calls on existing project notes in SKILL.md.
Boundary markers: None specified for the note ingestion process.
Capability inventory: write_note, npx skills add, uv tool install, and file writes to configuration files across the setup scripts.
Sanitization: No explicit validation or escaping of the ingested note content is performed before inference.

bm-setup