arkiv-write-safety
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides references to the source code for the @arkiv-network/sdk on unpkg.com, a well-known CDN, to facilitate implementation verification and auditing.
- [COMMAND_EXECUTION]: The provided JavaScript utility in references/resumable-import.md uses Node.js node:net to create a local mutex on 127.0.0.1 to prevent concurrent signer access, and node:fs for durable state management using restrictive 0o600 file permissions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes externally provided row data and payloads which could theoretically contain instructions intended to influence the agent's behavior during transaction preparation or result reconciliation.
- Ingestion points: The prepareImport function in references/resumable-import.md accepts a rows array containing payloads and attributes from external sources.
- Boundary markers: The skill instructions do not define specific prompt delimiters to isolate untrusted data from the agent's core instructions.
- Capability inventory: The skill possesses the capability to perform local filesystem writes (node:fs) and initiate blockchain transactions through a provider request interface.
- Sanitization: The implementation includes robust programmatic sanitization, including SHA256 integrity chains for journals, plan hashing, strict payload size limits, and schema validation.
Audit Metadata