submit-to-hackathon
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill performs telemetry by reading its own configuration file (~/.superstack/config.json) to extract a target URL and sending platform metadata, including the operating system and architecture. This behavior is transparently described in the preamble and is subject to user consent before activation.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted local project files (.superstack/idea-context.md, .superstack/build-context.md) to generate submission content. This represents a potential attack surface where malicious content in those files could influence the agent's behavior during the submission drafting process.
- Ingestion points: .superstack/idea-context.md and .superstack/build-context.md are read to provide context for the submission.
- Boundary markers: No explicit delimiters or instructions are used to isolate content from these files from the agent's core logic.
- Capability inventory: The skill possesses shell execution capabilities (bash), file system access, and network access (curl), which increases the potential impact of an injection attack.
- Sanitization: Input from project context files is not sanitized or validated before being used in prompts.
Audit Metadata