build-daily-inspiration-sites
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from
manifest.jsonandcontent.mdfiles to generate instructions for five new agent tasks. This creates a surface where malicious instructions embedded in the inspiration captures could influence the agent's behavior during landing page generation.\n - Ingestion points: Data is ingested from user-specified or workspace directories containing
manifest.jsonandcontent.md.\n - Boundary markers: The skill provides extensive instructions to "preserve the boundary," treat source content as art direction only, and ensure all identity and copy are replaced with original alternatives.\n
- Capability inventory: The skill reads local files, executes a bundled Node.js validation script, and initiates project builds and private deployments using platform tools (
sites).\n - Sanitization: The
validate_capture.mjsscript verifies manifest structure and file existence but does not perform content filtering or instruction detection on the textual fields.
Audit Metadata