qa-tdd-orchestrate

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from arguments and external files to drive the TDD orchestration process.
  • Ingestion points: Processes feature descriptions, acceptance criteria, and test cases provided in $ARGUMENTS or read from $JAAN_CONTEXT_DIR/tech.md.
  • Boundary markers: Employs a multi-level isolation architecture including artifact-only handoffs, prompt exclusion lists, and handoff manifest verification. A "HARD STOP" human review is required before Phase 2 execution.
  • Capability inventory: Spawns sub-agents via the Task tool, writes files to the $JAAN_OUTPUTS_DIR, modifies configuration files (settings.yaml), and executes shell commands for testing.
  • Sanitization: Restricts data transfer between agents to specific JSON manifests (handoff-red.json, handoff-green.json) containing file paths and test runner results.
  • [DYNAMIC_EXECUTION]: Executes shell commands for test running that are determined at runtime based on the detected project environment.
  • The specific test runner and command are resolved by reading local configuration files (tech.md) or via user input if detection fails.
  • Sources local library scripts (id-generator.sh, index-updater.sh) from the plugin root during the output generation and indexing phase.
  • [COMMAND_EXECUTION]: The skill executes shell-based test runners (e.g., vitest, jest, pytest) as part of the automated TDD cycle across RED, GREEN, and REFACTOR phases.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 09:30 AM
Security Audit — agent-trust-hub — qa-tdd-orchestrate