page-flow-cleanup-audit
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute standard development commands, such as
package build,package typecheck,git diff, andripgrep(rg), to validate refactoring results and search for code residues. These are legitimate operations for a code auditing tool. - [PROMPT_INJECTION]: The skill processes source code, which represents an indirect prompt injection surface. The analysis found no evidence of exploitation, and the structured workflow mitigates risks by focusing on logic rather than comments.
- Ingestion points: Source code files, project manifests, and type definitions found in the target repository.
- Boundary markers: The instructions do not define specific delimiters for separating code content from agent instructions.
- Capability inventory: The agent can read files and execute shell commands for building, testing, and searching code.
- Sanitization: Source code is processed without specific filtering or sanitization steps.
Audit Metadata