The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx -y @swarmvaultai/cli to download and execute the vendor's command-line interface. This is the standard method for using the tool provided by the author.
[COMMAND_EXECUTION]: The agent is instructed to use shell commands (swarmvault or npx) to perform operations such as compiling the vault, linting, and querying.
[PROMPT_INJECTION]: The skill processes external data (PDFs, emails, URLs) through its ingestion and compilation workflows, which is an inherent feature of knowledge management tools.
Ingestion points: External sources are added via ingest_input or by placing files in the raw/ directory.
Boundary markers: No explicit instructions are provided to the agent to treat ingested data as untrusted or to ignore instructions contained within it.
Capability inventory: The skill relies on file system access, shell execution, and network retrieval capabilities.
Sanitization: There are no mentioned mechanisms for sanitizing or filtering instructions from ingested files during the compilation or query process.

swarmvault