improve-animations
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains a defensive instruction (Hard Rule 4) directing the agent to ignore and flag any prompt injection attempts (e.g., "ignore previous instructions") encountered while reading repository files. The static detector flag for this category is a false positive resulting from this security-focused rule.\n- [INDIRECT_PROMPT_INJECTION]: The skill addresses potential indirect prompt injection by mandating that all repository content be treated as inert data. It includes a mandatory evidence chain evaluation:\n
- Ingestion points: Codebase files read during recon and audit phases.\n
- Boundary markers: Explicit instruction in Hard Rule 4 to treat file contents as inert and ignore steering attempts.\n
- Capability inventory: Limited to read-only operations (grep, git rev-parse) and writing plans to a specific plans/ directory.\n
- Sanitization: The agent is instructed to flag rather than execute instructions found in data.\n- [COMMAND_EXECUTION]: The skill uses standard, read-only version control and search tools (git, grep) for codebase analysis. It includes a strict prohibition against mutating operations, installs, or builds (Hard Rule 2).
Audit Metadata