ad-script-writing
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructional content for a creative writing task. It contains no executable code, remote resource fetching, or commands that could be exploited for privilege escalation or data exfiltration.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from project brief files, creating an attack surface for indirect prompt injection.\n
- Ingestion points: Untrusted content is read from
projects/<id>/brief.jsonandcreative/treatment.jsoninSKILL.md(Workflow Step 1).\n - Boundary markers: The skill relies on rigid JSON output structure and instructional constraints to prevent obedience to embedded commands, though explicit input delimiters are not used.\n
- Capability inventory: The skill is limited to reading project-specific metadata and writing a JSON script file to
projects/<id>/script/script.json(Workflow Step 7).\n - Sanitization: Security risk is mitigated by a mandatory validation step (
checkScript) and a separatead-compliance-reviewagent that inspects the output for legal and safety violations.
Audit Metadata