webcraft-copy
Warn
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill is instructed to fetch external copywriting guidelines from a third-party GitHub repository (github.com/coreyhaines31/marketingskills) during execution.
- Evidence:
references/upstream.mddirects the agent to resolve the latest commit and fetchskills/copywriting/SKILL.mdand related references. - [INDIRECT_PROMPT_INJECTION]: The skill incorporates instructions from an external source that is not controlled by the user or a trusted vendor, creating a vulnerability to prompt injection via the remote content.
- Ingestion points:
references/upstream.mdspecifies the remote repository and file path for the instructions. - Boundary markers: The skill includes specific rules in
references/upstream.mdstating that user instructions and local artifacts (e.g., the project brief) take precedence over any fetched guidance. - Capability inventory: The skill has the capability to read project metadata, draft and revise content in the file system (
03-copy.md), and recommend project-specific tool invocations (/webcraft-*). - Sanitization: There is no validation or filtering of the content fetched from the remote repository before it is processed as instructions by the agent.
Audit Metadata