everything-claude-code
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by instructing the agent to process various repository files and user-generated feedback.
- Ingestion points: The agent is guided to read and process content from repository files such as
SKILL.md,agents/*.md,commands/*.md, and review feedback. - Capability inventory: The skill workflows involve writing to the filesystem and generating executable code (
.sh,.js). - Boundary markers: There are no explicit instructions providing delimiters or warnings to ignore potentially malicious instructions embedded in the external repository data being processed.
- Sanitization: The skill does not outline specific validation or sanitization requirements for the content it ingests from the repository.
- [DYNAMIC_EXECUTION]: The skill describes workflows that involve the dynamic creation and modification of scripts within the repository environment.
- Evidence: The "Add New Skill" and "Add Or Update Hook" workflows explicitly guide the agent to create and modify shell (
.sh) and JavaScript (.js) scripts. While this is an intended feature for repository extension, it constitutes a dynamic code generation capability.
Audit Metadata