The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill repeatedly invokes bunx @plaited/development-skills. This command downloads and executes a package from the npm registry at runtime.
Evidence: The scripts lsp-hover, lsp-symbols, lsp-refs, lsp-find, and lsp-analyze all rely on executing code from the @plaited organization.
Trust Status: The @plaited organization is not included in the list of 'Trusted GitHub Organizations' or 'Trusted GitHub Repositories'. Per the analysis framework, execution of unverifiable packages from non-trusted sources is classified as MEDIUM severity.
[DATA_EXPOSURE] (SAFE): The skill accesses local source code files (*.ts, *.js, etc.) to provide LSP functionality. There is no evidence of attempts to access sensitive system files, environment variables, or credentials.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted external data (source code files) and produces structured JSON output for an agent. While this creates a potential surface for injection if an attacker-controlled file is scanned, the risk is inherent to the tool's purpose and the output is structured data.
Ingestion points: Processes files matching *.ts, *.tsx, *.js, *.jsx via LSP tools.
Boundary markers: Output is formatted as JSON, which provides basic structural delimiters.
Capability inventory: Uses Bash via bunx to run analysis scripts.
Sanitization: No specific sanitization or 'ignore instructions' markers are mentioned in the skill definition.

typescript-lsp