handoff
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The troubleshooting documentation instructs the agent to read secrets from sensitive local paths (~/.agentmemory/.env and ~/.agentmemory/secret) as a fallback when the primary memory daemon connection fails. While the instructions provide safety constraints (e.g., using loopback URLs and prohibiting display of the secret), accessing local credential files is a sensitive behavior.
- Evidence: File:
TROUBLESHOOTING.md, text: "fall back to the secret in ~/.agentmemory/.env and then ~/.agentmemory/secret." - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the memory store, including session summaries and conversation narratives, to facilitate session resumption. This content could potentially include malicious instructions from previous sessions.
- Ingestion points:
memory_sessionsandmemory_recalltool outputs processed inSKILL.md. - Boundary markers: None. The recalled content is summarized and integrated directly into the agent's context.
- Capability inventory: The skill uses
memory_sessionsandmemory_recalltools to query and retrieve historical session data. - Sanitization: No explicit sanitization or instruction to ignore embedded directives is present in the workflow.
Audit Metadata