The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the acquisition and installation of external packages and CLI tools from the inference.sh and belt-sh registries.
Evidence: instructions provided for npx skills add belt-sh/cli and various sub-packages under inference-sh/skills.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to manage skill lifecycles. Security risk is mitigated by the allowed-tools frontmatter configuration, which restricts execution to the npx skills command space.
Evidence: allowed-tools: Bash(npx skills *) in the YAML frontmatter.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface by processing user-provided search terms and installing content from a remote registry which could theoretically contain instructions intended to influence the agent.
Ingestion points: User-supplied search queries via npx skills search [query].
Capability inventory: Restricted Bash access limited to skill management commands.
Boundary markers: Not explicitly defined in the instruction text.
Sanitization: Relies on the underlying CLI tool and agent environment for input validation.

related-skill