The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation recommends installation via standard package managers and tool runners such as npm, apm, and uvx, which fetch content from well-known repositories (GitHub) and registries (PyPI).
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided SQL files, making it a surface for indirect prompt injection.
Ingestion points: SQL file content is ingested by the scripts/catalog-lint.mjs utility using fs.readFileSync based on CLI arguments provided by the agent or user.
Boundary markers: There are no explicit markers wrapping the read data to distinguish it from instructions.
Capability inventory: The linter script is limited to read-only file system access and does not have the capability to perform network requests, write files, or execute arbitrary shell commands.
Sanitization: The tool performs static analysis using regular expressions and only outputs specific metadata and pre-defined error messages, preventing raw input from being interpreted as instructions by the agent.
[SAFE]: The provided linter script is a simple, dependency-free JavaScript file with transparent logic. No obfuscation, persistence, or data exfiltration patterns were detected during the analysis.

sql-lint