memory-graph-builder
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes the
MEMORY.mdfile to generate amemory-digest.mdfile, which is explicitly intended for injection into the agent's system prompt. If the memory file contains malicious instructions from previous interactions, these could be propagated into the agent's core context. - Ingestion points: The skill reads from
~/.openclaw/MEMORY.mdin theparse_memory_filefunction. - Boundary markers: The generated digest does not wrap memory entries in distinct security boundaries or provide instructions to the agent to ignore embedded commands within the memory content.
- Capability inventory: The skill has the ability to read from the filesystem (
MEMORY.md) and write to both the skill state (state.yaml) and the workspace (memory-digest.md). - Sanitization: The script performs basic markdown cleaning and keyword-based classification but does not sanitize the actual content of the memory entries before including them in the digest.
Audit Metadata