claude-skill-orchestration-audit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content by reading SKILL.md files from the local filesystem and passing the raw text to subagents for analysis.
- Ingestion points: Files located in ~/.claude/skills/*/SKILL.md or a user-specified directory are read and their content is used as input for subagents.
- Boundary markers: The instructions in references/agent-skill-analyzer.md do not define delimiters (such as XML tags or triple backticks) to separate the skill content from the analysis instructions, increasing the risk that malicious instructions within an audited skill could hijack the subagent's task.
- Capability inventory: The skill has the ability to launch subagents, read files, and write to the filesystem via the --fix parameter which uses the Edit tool.
- Sanitization: There is no evidence of sanitization or filtering applied to the skill content before it is processed by the AI subagents.
Audit Metadata