The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs three third-party packages globally from the npm registry: @steipete/summarize, @tobilu/qmd, and agent-browser. These packages are not from trusted organizations and are installed without version pinning, which introduces potential risk from the software supply chain.
[COMMAND_EXECUTION]: The skill executes a bundled shell script (scripts/onboarding.sh) to scaffold the vault and verify environment state. It also invokes the npm CLI to install global tools and the agent-browser install command to set up local dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process raw source documents (clipped web articles or files) into a structured wiki. This workflow processes external, potentially untrusted content through the agent's logic.
Ingestion points: Raw files added to the raw/ directory and processed via the /second-brain-ingest flow (referenced in SKILL.md and wiki-schema.md).
Boundary markers: The instructions do not define specific delimiters or warnings to ignore embedded instructions within the raw source material.
Capability inventory: The agent has access to Bash, Read, Write, Glob, and Grep tools and executes both bundled and downloaded CLI tools.
Sanitization: There is no evidence of input validation or sanitization applied to the content extracted from raw source files during the ingestion process.

second-brain