adr-index
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Node.js script located at
plugins/ruflo-adr/scripts/import.mjsusing the Bash tool. This script is part of the skill's own package and is used to automate the indexing process, which is more efficient than individual tool calls. - [INDIRECT_PROMPT_INJECTION]: The skill parses content from files in the repository's ADR directories and writes that data to a persistent memory store, creating a surface area for processing untrusted data.
- Ingestion points: Files matching the patterns
*/docs/adr/or*/docs/adrs/on the local filesystem. - Boundary markers: No explicit delimiters or instructions are used to separate the content being indexed from the agent's internal control logic.
- Capability inventory: The skill utilizes the Bash tool for script execution and memory tools (
mcp__plugin_ruflo-core_ruflo__memory_list,mcp__plugin_ruflo-core_ruflo__memory_search) for data persistence. - Sanitization: The skill includes a regex-based 'false-positive guard' to strip issue and PR references before extraction to prevent metadata corruption, though it does not explicitly describe sanitization against embedded instructions.
Audit Metadata