memori
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from external code repositories, which presents an attack surface for indirect prompt injection if repository files contain malicious instructions aimed at the agent.
- Ingestion points: The skill reads files from the local filesystem (e.g., README.md, source code) using the Read tool and the preflight.py script.
- Boundary markers: Skill instructions emphasize inspection and require explicit approval before performing actions based on the code.
- Capability inventory: The skill has access to Bash, which could theoretically be used for command execution if the agent is influenced by malicious repository content.
- Sanitization: The preflight.py script implements path validation and blocks access to sensitive filenames and directories (e.g., .ssh, .aws, .env).
- [COMMAND_EXECUTION]: The skill provides a local Python script for repository verification.
- Evidence: SKILL.md instructs the agent to run 'python3 scripts/preflight.py'.
- Analysis: The script uses only standard library modules and performs no network or subprocess calls. It implements secure file traversal using POSIX file descriptors to prevent directory traversal or symlink-based attacks.
- [EXTERNAL_DOWNLOADS]: The skill references an external repository and project metadata.
- Evidence: References 'https://github.com/MemoriLabs/Memori' and suggests installation of the 'memori' package.
- Analysis: These are legitimate resources for the skill's purpose. The skill includes a hashing mechanism to verify that the target content matches the expected versions.
Audit Metadata