short-drama-knowhow
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a qualitative research tool that ingests external textual data to distill production knowledge into public rubrics and references.
- Ingestion points: The agent reads full project chains including creative briefs, story structures, screenplays, and revision records as defined in
SKILL.mdandreferences/prompt-production-learning.md. - Boundary markers: The instructions require establishing a "read-only authorization" and an "isolated workspace" (
SKILL.md, Step 1). However, specific prompt delimiters for separating ingested text from instructions are not explicitly defined in the provided schema files. - Capability inventory: The skill is authorized to perform file reads, comparisons, and invoke existing tools for deduplication and privacy scanning (
SKILL.md, Core Principles). - Sanitization: The skill mandates a multi-stage process for de-identification and "de-copying" of content, followed by a blind forward evaluation by a fresh agent and review by an independent human/agent (
references/synthesis-and-promotion.mdandreferences/blind-forward-eval.md).
Audit Metadata