simplify
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to operate autonomously and proactively without explicit user requests, which reduces human oversight during code modifications.
- [EXTERNAL_DOWNLOADS]: The skill references http://CLAUDE.md as a source for coding standards. Because .md is a valid top-level domain, this URL could resolve to an external source rather than the intended local project file.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted code content without sufficient boundary markers. 1. Ingestion points: Recently modified code files in the active session. 2. Boundary markers: Absent; the instructions lack explicit delimiters or warnings to ignore instructions embedded in the code. 3. Capability inventory: Implicit file system write access used for code refinement. 4. Sanitization: Absent; no validation or filtering of ingested code is performed.
Audit Metadata