security-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The 'Overrides' section in
SKILL.mdexplicitly instructs the agent to prioritize and follow specific instructions found within the project's own documentation and prompt files, even when they require bypassing established security best practices. This provides a direct path for malicious instructions embedded in untrusted project data to override the agent's security logic.\n - Ingestion points: The skill directs the agent to 'inspect the repo to determine it' and to 'pay attention to specific rules and instructions in the project's documentation and prompt files' as described in
SKILL.md.\n - Boundary markers: The instructions lack any requirement for boundary markers or validation of the override instructions, effectively granting untrusted codebase content the authority to dictate security policy.\n
- Capability inventory: The agent typically utilizes core file system and command execution tools to perform reviews and apply fixes, which could be leveraged to implement insecure code if triggered by a prompt injection.\n
- Sanitization: There is no instruction to sanitize or verify the legitimacy of these override requests before implementation.
Audit Metadata