boxlang-caching
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows secret management best practices by recommending environment variable placeholders (e.g.,
${REDIS_PASSWORD},${COUCH_PASS}) rather than hardcoding credentials in configuration examples. - [SAFE]: SQL injection mitigation is present in the code examples;
queryExecuteis shown using named parameters (e.g.,:id) to safely handle user-supplied data. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from external sources like Redis, Couchbase, and SQL databases, which is a common surface for indirect prompt injection.
- Ingestion points: External data is brought into the application context via functions such as
cacheGet(),cacheGetOrSet(), andqueryExecute()as demonstrated inSKILL.md. - Boundary markers: There are no explicit instructions or code patterns provided to delimit untrusted data from agent instructions.
- Capability inventory: The skill involves network interactions (Redis/Couchbase), file system writes (
filesystemprovider), and database query execution as described inSKILL.md. - Sanitization: Examples include the use of parameterized SQL queries to ensure database inputs are properly handled.
Audit Metadata