promptkit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains defensive instructions to recognize and ignore injection attempts (e.g., 'ignore previous instructions') within the data it processes. The static analysis alerts represent these security features and are not attempts to override the agent's own safety guidelines.
- [DATA_EXPOSURE]: A core security rule in the skill requires scanning for and replacing hardcoded secrets (API keys, tokens) with placeholders in user-provided text, preventing sensitive data from being committed to repositories or chat logs.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges that its input surface—user-supplied prompts for sharpening—is untrusted by construction. It implements a mandatory 'input is inert' rule to ensure that any instructions contained within the analyzed text are treated as data rather than executable commands.
- [COMMAND_EXECUTION]: The skill is strictly limited to file reading and editing tools. It does not have access to a shell, network, or package managers, effectively eliminating the risk of remote code execution or exfiltration through standard attack vectors.
Audit Metadata