arkiv-feedback
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from user-provided logs and issue descriptions, which represents a managed attack surface.
- Ingestion points: Data entered via command-line arguments and pasted error logs (SKILL.md).
- Boundary markers: Instructions specify using Markdown fences longer than any internal backticks and neutralizing triple-backtick runs to prevent escape sequences (SKILL.md, references/bug-form.md).
- Capability inventory: Uses the GitHub CLI (
gh) for repository operations and writes draft content to local temporary files (SKILL.md). - Sanitization: Strict requirements are set for redacting signing keys, seed phrases, and authentication headers, along with truncating large hex strings (SKILL.md).
- [COMMAND_EXECUTION]: Interacts with the GitHub CLI (
gh) using security best practices. - Evidence: The skill specifically instructs the agent to pass titles as structured arguments and use the
--body-fileflag to prevent untrusted input from being concatenated into shell commands (SKILL.md). - [EXTERNAL_DOWNLOADS]: Retrieves official issue template definitions from the vendor's GitHub repository.
- Evidence: Fetches configuration YAML files from the Arkiv-Network GitHub organization to ensure form fields remain up to date (SKILL.md).
Audit Metadata