assimilate

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from untrusted external data.\n
  • Ingestion points: Untrusted source code is cloned from external repositories into the .claude/context/runtime/assimilate/ workspace during Phase 1.\n
  • Boundary markers: The skill lacks explicit boundary markers or instructions to treat external code as untrusted data during analysis.\n
  • Capability inventory: The skill generates a TDD backlog involving Write, Edit, and Bash commands.\n
  • Sanitization: There is no mechanism to sanitize or validate the content of cloned repositories before it influences the generated implementation plans.\n- [PROMPT_INJECTION]: The skill metadata contains a verified: true claim. This is a self-reported status by the author and does not represent a system-level security guarantee.\n- [EXTERNAL_DOWNLOADS]: The skill clones external repositories for analysis purposes.\n
  • Evidence: Phase 1 uses git clone --depth=1 to fetch remote codebases. This risk is mitigated by 'Iron Laws' forbidding the execution of scripts (e.g., npm install, make, ./setup.sh) from these sources.\n- [COMMAND_EXECUTION]: The skill performs command-line operations for framework management.\n
  • Evidence: Executes TOOL --help for CLI discovery and pnpm skills:index to update the skill registry.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 12:48 PM