claude-code-learning
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a setup feature that analyzes project-specific files like CLAUDE.md and local configuration files in the .claude directory. This creates an attack surface where untrusted data within a project could potentially influence the agent's behavior during the setup process.
- Ingestion points: CLAUDE.md, .claude/settings.local.json, .mcp.json, .claude/commands/, .claude/agents/, .claude/skills/.
- Boundary markers: None identified; the skill does not explicitly use delimiters to separate ingested project content from its own instructions.
- Capability inventory: The skill uses Read, Write, Edit, Glob, Grep, Bash, and AskUserQuestion tools.
- Sanitization: No specific sanitization or validation logic for the ingested configuration data is described in the skill instructions.
- [SAFE]: The skill's access to local configuration files, including sensitive paths like .mcp.json, is consistent with its stated purpose of managing and optimizing the Claude Code environment. There is no evidence of these files being exfiltrated to external domains or used for unauthorized command execution.
Audit Metadata