memory-management

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill architecture facilitates the collection of sensitive organizational and personal data into a predictable directory structure (memory/ and CLAUDE.md). This includes personnel roles, project codenames, budget details, and user preferences. While no network exfiltration is defined, this centralized data exposure increases the impact of a potential compromise of the agent or the local environment.
  • [PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface (Category 8) by design. It instructs the agent to 'Always decode shorthand before acting' based on content found in external files, which could contain malicious instructions.
  • Ingestion points: Content is ingested from CLAUDE.md, memory/glossary.md, and various markdown files within the memory/ subdirectory (SKILL.md).
  • Boundary markers: The skill lacks requirements for boundary markers or delimiters when reading memory files, making the agent more likely to follow instructions embedded within the data.
  • Capability inventory: The skill utilizes file system read and write operations to manage and recall memory (SKILL.md).
  • Sanitization: No sanitization, validation, or escaping of the ingested markdown content is specified before the agent processes it for logic decoding.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:21 PM
Security Audit — agent-trust-hub — memory-management