architecture-md-builder

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell commands and scripts to perform its tasks.
  • It calls python3 ~/.claude/skills/exa-search/scripts/exa_search.py and exa_research.py to conduct background research on documentation best practices.
  • It uses wc -l to calculate line counts for source files as part of the architecture mapping process.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the repository being documented.
  • Ingestion points: Phase 2 exploration agents read entry points, abstractions, data structures, and API handling code from the local codebase (SKILL.md).
  • Boundary markers: The subagent prompts do not include explicit delimiters or instructions to ignore instructions embedded within the source code being analyzed.
  • Capability inventory: The subagents are tasked with mapping core systems, data flows, and architectural invariants, providing a wide scope for influence (SKILL.md).
  • Sanitization: There is no evidence of sanitization or filtering of source code content before it is processed by the exploration agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 11:39 PM