fable-goal
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text ('rambles') to create prompts for subsequent agent sessions that include directives for complete autonomy ('do not ask me for anything until you are all done') and broad resource discovery mandates. This creates a surface for instructions that could influence the behavior of an autonomous agent in a downstream session. As the output is a text code block intended for the user to review and manually paste, the risk of automated exploitation is low and inherent to the tool's purpose.
- Ingestion points: User-provided input interpreted as a 'ramble' (SKILL.md).
- Boundary markers: None; the skill is instructed to weave the user's intent into a natural prose prompt template without delimiters.
- Capability inventory: The generated prompts encourage the use of iteration passes and tool discovery, including fetching references or assets (SKILL.md, references/goal_prompt_patterns.md).
- Sanitization: The skill does not perform sanitization on user input before integrating it into the prompt template.
Audit Metadata