The Agent Skills Directory

[DYNAMIC_EXECUTION]: The TypeScript code examples demonstrate the use of child_process.execSync to run CLI commands by interpolating variables directly into shell strings (e.g., vs search run --app-id ${appId} --query "${query}" --json). If an agent or user implements this pattern without strict sanitization of the appId or query variables, it could lead to arbitrary command execution via shell metacharacters.- [INDIRECT_PROMPT_INJECTION]: The skill processes external data files (e.g., items.json, products.json) to generate provisioning plans and perform automated search quality tuning with LLM-based judging.
Ingestion points: Data enters via vs item profile --file ./items.json and vs search tune plan --queries @queries.json.
Boundary markers: None identified in the skill instructions or examples.
Capability inventory: The skill uses execSync for shell commands and interacts with remote APIs.
Sanitization: No explicit sanitization or validation of the input data content is mentioned before it is processed by the agent or the tuning LLM.- [EXTERNAL_DOWNLOADS]: The installation instructions involve cloning a repository from GitHub (github.com/volcengine/SearchCLI.git) and executing a local shell script (bash ./scripts/install.sh). This fetches and executes code from a remote service provider.

searchcli-volcengine-ai-search