developer-community-moderation
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The analysis of the skill's instructions, reference materials, and evaluation prompts did not identify any malicious patterns or vulnerabilities. The skill adheres to security best practices and leverages established standards from the open-source community, such as the Contributor Covenant and Django moderation guidelines. All external references point to trusted organizations or well-known public resources.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests untrusted data from community members (e.g., incident reports, existing code of conduct files) to generate official moderation policies and runbooks. \n
- Ingestion points: Data is gathered during the interview phase (Question 8 and 9 in SKILL.md) and through incident descriptions provided in the evaluation prompts (evals/evals.json). \n
- Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore instructions' warnings when processing user-provided incident details. \n
- Capability inventory: The skill is restricted to generating text-based deliverables (Markdown files) and does not utilize tools for file system modification, network communication, or arbitrary command execution. \n
- Sanitization: No specific sanitization or validation protocols for user-provided input are defined prior to the data being interpolated into policy drafts. \n
- Note: This surface is considered safe given the skill's limited capabilities and its intended use case for policy drafting.
Audit Metadata