ragcheck
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute a local CLI binary named
ragcheckto perform its primary functions.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install theragchecktool via Homebrew (itamaker/tap/ragcheck) or by downloading releases from the author's GitHub repository (github.com/itamaker/ragcheck-skill). These resources are provided by the skill's author and represent standard installation procedures for the utility.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data provided in JSON format (such as questions, answers, and contexts) which serves as an ingestion surface.\n - Ingestion points: Inputs enter the agent context via the
-qrels,-run, and-inputflags as described inSKILL.mdandREFERENCE.md.\n - Boundary markers: The data is handled in structured JSON arrays.\n
- Capability inventory: The tool performs offline lexical analysis and prints reports to the console; it does not perform network operations or model calls during this process.\n
- Sanitization: The tool uses a deterministic tokenization process (lowercasing, filtering short tokens, and removing stopwords) as documented in
REFERENCE.md, minimizing the risk of instructions being interpreted as code.
Audit Metadata