loop-circuit-breaker
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes tool failure data, including error messages and arguments, which could potentially contain malicious instructions if a tool interacts with untrusted external content.
- Ingestion Points: The
check.pyscript receives error strings and tool arguments via command-line flags (SKILL.md,check.py). - Boundary Markers: No specific delimiters or safety warnings are provided to the agent for parsing these errors.
- Capability Inventory: The skill can write to a local state file and instructs the agent to decide whether to stop or continue the task based on the error content (
SKILL.md). - Sanitization: Error messages are truncated to 200 characters in the state file, but no content filtering or escaping is performed (
check.py). - [COMMAND_EXECUTION]: The skill uses a local Python script to manage its state and detection logic. This is an intended and safe use of local execution for project management.
- [SAFE]: The skill maintains state in a dedicated local directory (~/.openclaw/skill-state/) and does not request network access or perform any obfuscated operations.
Audit Metadata