skills/afelipeg/anthropic-skills-for-enterprise-marketing-os/client-memory-synthesizer/Gen Agent Trust Hub
client-memory-synthesizer
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by design. Its primary purpose is to aggregate and synthesize external data—such as SOWs, email threads, and meeting notes—into a 'Tenant Context' which serves as persistent Project Instructions for the AI agent. If these source documents contain malicious instructions, they could be promoted into the agent's core operational logic.\n
- Ingestion points: Untrusted data enters the system through the
memory_inputsdictionary inscripts/memory_synthesizer.py, which populates fields like business objectives, commercial scope, and operating rules from external sources.\n - Boundary markers: The
_generate_tenant_contextfunction inscripts/memory_synthesizer.pyinterpolates input data directly into markdown headers without implementing boundary markers or explicit instructions for the AI to ignore embedded commands.\n - Capability inventory: While the synthesizer script itself only performs data processing and file I/O, the resulting 'Tenant Context' defines the operating environment and guardrails for all other skills active in the project workspace.\n
- Sanitization: There is no evidence of input validation, sanitization, or escaping of the strings provided in the
memory_inputsbefore they are formatted into the final instruction prompt.
Audit Metadata