compressed-json-repair
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides documentation and code snippets for local JSON healing. The logic uses standard regular expressions and the well-known
jsonrepairlibrary to fix structural formatting errors without performing dangerous operations like file system writes or network requests. - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted data (model-generated strings).
- Ingestion points: The
raw_textparameter in both the Python and TypeScript functions (SKILL.md). - Boundary markers: None identified in the provided repair snippets.
- Capability inventory: The functions are limited to string manipulation and standard parsing (
json.loads,JSON.parse). No network, file, or shell execution capabilities are present. - Sanitization: The logic attempts to sanitize structural errors but does not validate the semantic content of the JSON. Given the lack of dangerous capabilities, this surface is considered safe.
Audit Metadata