codex-thread-orchestrator

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The hooks/user-hooks.json file configures a SessionStart hook that executes a local script, scripts/compact_rehydration.py, when a session is initiated with the 'compact' matcher. This script is used to inject the file paths of durable state back into the agent's context.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads external state files and instructs the agent to treat them as a 'durable source of truth.'
  • Ingestion points: scripts/compact_rehydration.py reads control.md and child markdown files within the .agents/workflows/ directory to resolve session-to-file mappings.
  • Boundary markers: The script provides explicit instructions to the agent to re-read the specific file path, effectively using the file content as a state override.
  • Capability inventory: The orchestrator utilizes thread management tools like create_thread and wait_threads to delegate work to sub-tasks.
  • Sanitization: Input validation is performed on the thread_id using a regex (SAFE_ID) that restricts characters to alphanumeric and safe symbols, preventing path traversal attacks during file lookup.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 10:11 PM
Security Audit — agent-trust-hub — codex-thread-orchestrator