no-leaky-app
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest HTTP headers from a web application, which provides a surface for indirect prompt injection if those headers contain malicious instructions.
- Ingestion points: In
SKILL.md(Step 1), the agent is instructed to collect headers from application URLs using `curl -s -D - -o /dev/null URL`.
- Boundary markers: The skill does not define any delimiters or provide instructions to the agent to treat the header content as untrusted data.
- Capability inventory: In
SKILL.md(Step 3), the agent is given the capability and instruction to modify the application source code to fix header configurations. - Sanitization: The skill lacks mechanisms to sanitize or validate the ingested header strings before the agent processes them.
Audit Metadata