deslop
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute Python scripts (
slop_lint.pyandverify_edit.py) located in sibling directories. These are used for text analysis and to verify that quotes, code, and numbers remain unchanged during the editing process. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (draft text) which is used as data for the linting and verification tools.
- Ingestion points: User-provided drafts are saved to local files (
original.md) and processed by the scripts. - Boundary markers: The workflow requires the use of a JSON-formatted linting report to identify specific 'spans' for editing, providing a logical boundary that prevents the agent from rewriting unflagged content.
- Capability inventory: The skill utilizes subprocess execution of local Python tools for content analysis and integrity checking.
- Sanitization: The skill incorporates a mandatory verification step (
deslop-verify) which checks for hard-invariant failures (e.g., changes to quotes or numbers) and requires the agent to revert changes if these invariants are violated.
Audit Metadata