atlas-obscura
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The script
ao.mjsusesexecSyncto automatically runnpm installif it detects that theatlas-obscura-apidependency is missing from the local environment. While intended for automated setup within the skill's own directory, this constitutes local command execution. - [INDIRECT_PROMPT_INJECTION]: The skill fetches place descriptions, directions, and tags from the external Atlas Obscura website, creating a potential attack surface for indirect prompt injection.
- Ingestion points: External content is retrieved via
atlasObscura.searchandatlasObscura.placeFullinao.mjs. - Boundary markers: None identified; the skill outputs structured JSON directly.
- Capability inventory: The script includes
execSyncfor package installation and file system access to check for existing modules. - Sanitization: The skill performs basic sanitization by stripping HTML tags and entities using regular expressions (e.g.,
replace(/<[^>]*>/g, '')) before data is returned. - [EXTERNAL_DOWNLOADS]: The skill downloads the
atlas-obscura-apipackage and its associated dependency tree from the official NPM registry as part of its installation and first-run process.
Audit Metadata