gathering-migration

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests codebase data and user instructions which are passed to subagents, creating a potential injection surface.
  • Ingestion points: SKILL.md receives the migration request, and references/conductor-dispatch.md receives the scouted territory_map derived from codebase analysis.
  • Boundary markers: Variables {migration_spec} and {territory_map} are interpolated into prompts in references/conductor-dispatch.md without the use of strict delimiters or instructions to the model to ignore embedded directives.
  • Capability inventory: The workflow utilizes gf for codebase searching, gw for git operations, and executes pnpm install and test commands via the shell.
  • Sanitization: No sanitization or filtering logic for the ingested content is defined in the instructions.
  • [COMMAND_EXECUTION]: The skill instructs the agent to run shell commands for project maintenance and verification.
  • Evidence: Gate checks in SKILL.md (Phases 3 and 4) execute pnpm install and gw dev ci --affected --fail-fast --diagnose to verify the state of the codebase.
  • [EXTERNAL_DOWNLOADS]: The skill performs dependency resolution which involves downloading external assets.
  • Evidence: The pnpm install command is executed as part of the validation process to ensure the migration has not broken the dependency tree.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 01:31 AM