hackernews-playbook
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The skill consists of instructions and documentation. It does not contain any executable scripts or binary files, preventing any risk of unauthorized code execution.\n- [SAFE]: The guidelines provided by the skill are aligned with the official rules and community norms of Hacker News, emphasizing organic engagement and transparency while explicitly discouraging vote manipulation and deception.\n- [INDIRECT_PROMPT_INJECTION]: The skill addresses the risk of processing untrusted data (user-provided drafts) by explicitly instructing the agent to refuse requests to polish or generate prose for the user. It provides fictional templates to guide the user without directly interpolating or transforming the user's input into an output.\n
- Ingestion points: User-provided drafts in the conversation prompt as described in the 'Interview' and 'Output' sections of SKILL.md.\n
- Boundary markers: Explicit instructions to 'Never return finished HN-facing prose' and 'Refuse to polish or write the user's Show HN text' act as strong delimiters between user input and the agent's output.\n
- Capability inventory: The skill possesses no file system access, network operations, or subprocess capabilities, limiting the potential impact of any successful injection.\n
- Sanitization: The skill uses independent fictional examples (e.g., 'Litequeue', 'Fernwood Labs') to demonstrate structure and tone rather than processing or sanitizing user input directly.
Audit Metadata