lov-better-readme
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from workspace files (READMEs, manifests, and scripts) to generate updated documentation, which represents an inherent surface for indirect prompt injection.\n
- Ingestion points: The workflow in
SKILL.mdspecifies reading project root files, entry points, and script contents.\n - Boundary markers: No explicit delimiters or instructions are used to separate ingested project data from the agent's system prompt.\n
- Capability inventory: The skill can modify local files and execute git commands (commit, push) based on user instructions.\n
- Sanitization: No sanitization of ingested project content is performed before processing by the agent.\n- [EXTERNAL_DOWNLOADS]: The
scripts/validate_skill.pyutility referencesPyYAML, a standard YAML parsing library, which is typically installed from official package registries for developer validation tasks.\n- [COMMAND_EXECUTION]: The skill performs standard development operations, such as executing git commands to commit and push changes and running local project scripts to verify documentation accuracy, as defined in theSKILL.mdworkflow.
Audit Metadata