memory-clarity-probe
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external or session-derived text through the
{memory_content}variable. - Ingestion points: The variable
{memory_content}is interpolated into prompts within Step 2 and Step 3 of the Core Workflow inSKILL.md. - Boundary markers: The skill uses XML-style
<memory>and</memory>tags to delimit the untrusted content. - Capability inventory: The skill itself performs no high-risk operations; it does not write files, spawn subprocesses, or perform network requests.
- Sanitization: There is no explicit sanitization or escaping performed on the
{memory_content}before interpolation, which may allow an adversary to craft content that attempts to influence the agent's scoring of the probe.
Audit Metadata