debug-like-expert-skill
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions utilize standard shell commands such as find, ls, and grep to identify project file types and discover local expertise files, which is consistent with its function as a developer tool.
- [INDIRECT_PROMPT_INJECTION]: As a debugging utility, the skill processes untrusted user source code and external information from web searches or documentation tools. While it lacks explicit sanitization or boundary markers for this data, this is a standard risk for the tool's intended use case. Ingestion points include local source code files and web search results; capabilities include file reading, file writing for telemetry, and command execution for verification.
- [PERSISTENCE]: The skill implements a local logging mechanism by instructing the agent to write session outcomes to a JSON file located in a hidden application directory (~/.claude/skill-analytics/). This is used for internal analytics and does not involve external data transmission.
Audit Metadata