team-tech-debt

Pass

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to perform essential project maintenance tasks, such as creating isolated git worktrees via git worktree and running validation suites using npm test, npx tsc, npx eslint, and pytest. It also coordinates with an AI-driven platform tool ccw cli to execute code analysis and apply refactoring changes across the codebase.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes arbitrary repository content.\n * Ingestion points: Source code files and documentation are read by the scanner role across multiple dimensions (roles/scanner/role.md).\n * Boundary markers: The executor role (roles/executor/role.md) utilizes explicit prompt constraints—such as "Minimal changes only" and "No suppressions"—when instructing the code-writing utility to maintain control over the modification process.\n * Capability inventory: The skill leverages the Bash, Write, Edit, and Agent tools to modify code, manage worker lifecycles, and execute validation scripts.\n * Sanitization: No specific data sanitization or instruction-filtering mechanism is documented for the transition between reading codebase content and generating remediation prompts.\n- [SAFE]: The orchestration logic is implemented via a coordinator role that spawns specialized workers using the Agent tool. This modular design uses locally defined specifications to load domain logic for each phase of the pipeline (scan, assess, plan, fix, validate), ensuring that all automated actions are aligned with the skill's primary objective of debt remediation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 5, 2026, 04:03 AM