custom-agent-usage
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the discovery and ingestion of content from local
.agent.mdfiles. This behavior establishes a potential vector for indirect prompt injection if those external files are manipulated to include malicious instructions. - Ingestion points: Agent configuration files found in specific local directories (C:\Users\LOQ.claude\agents and C:\Users\LOQ\AppData\Roaming\Code
- Insiders\User\prompts) are parsed in SKILL.md and scripts/agent-finder.js.
- Boundary markers: The skill relies on standard YAML frontmatter delimiters (---) to separate metadata from instructions.
- Capability inventory: The skill provides the ability to delegate tasks to discovered agents via the runSubagent tool.
- Sanitization: Ingested data is processed via basic regular expressions in scripts/agent-finder.js without explicit filtering for executable or malicious prompt content.
Audit Metadata