Back to skill

Security audit

Codex Deepsearch

Security checks for vulnerabilities and agentic risk

Overview

This is a research-workflow skill that asks Codex to use evidence ledgers, privacy limits, and review gates, with no executable payload or hidden install behavior found.

Installers should expect this skill to guide Codex through structured research that may read local project files, browse current sources, create research documents, and optionally delegate work when permitted. Review the missing referenced materials if the publisher intended them to be included, but the inspected artifact does not show hidden execution or unsafe behavior.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · SKILL.md (reported line 29)May include surrounding context.

md
- Treat `AGENTS.md` and repo docs as maps and constraints, not as proof of external facts.
- Prefer `rg`, `sed`, `git status`, and `git diff` for local inspection.
- Use `apply_patch` for manual edits when the user explicitly asks to create research artifacts or update skill files.
- Do not default to writing inside a project repository. Ask or use an explicitly authorized research directory.
- Do not commit, stage, or push unless the user explicitly asks.
- If Python is needed, use `uv` or an existing project/skill validation command; do not modify system Python.