The Agent Skills Directory

[DATA_EXFILTRATION]: The skill configuration references ~/.neuralmemory/config.toml in its context. This is a tool-specific configuration file used to establish the baseline for the memory audit. There are no commands present that attempt to exfiltrate this data to external or non-whitelisted domains.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes data stored in "memories" via retrieval tools, which represents an ingestion surface for potentially untrusted content.
Ingestion points: Memory content is retrieved through nmem_recall and nmem_context calls specified in Phase 1 and Phase 2.
Boundary markers: The instructions do not define specific delimiters or "ignore" markers for the content being audited.
Capability inventory: The agent is restricted to a specific set of tools (nmem_recall, nmem_stats, nmem_health, nmem_context, nmem_conflicts) which are read-only or limited to specific memory management actions. No generic shell execution or network capabilities are permitted.
Sanitization: The skill relies on the agent's internal reasoning to identify "poisoned data" or "unverified claims" as part of its audit logic, but no programmatic sanitization of the input data is described.

memory-audit