remove-ai-slops
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs local shell operations such as "git diff" to identify modified files and "awk" with "wc" to calculate pure lines of code. These are appropriate uses of system tools for code analysis within the project's scope.
- [PROMPT_INJECTION]: The skill ingests project source code and includes it in prompts for sub-agents during the cleanup phase. This presents a potential surface for indirect prompt injection if malicious instructions are present in the files being processed. * Ingestion points: Files identified in the scope are processed by sub-agents in Phase 4. * Boundary markers: The prompt template for sub-agents does not utilize robust delimiters to separate instructions from the file content. * Capability inventory: The agents have the ability to modify the filesystem and execute shell commands (tests/linting). * Sanitization: The skill relies on functional verification (Phase 2 regression tests and Phase 5 quality gates) to ensure that only intended slop removal occurs, providing a strong defense-in-depth against malicious logic changes.
Audit Metadata