remember-context
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill creates a mechanism for persisting untrusted user input into files that define the agent's future behavior and constraints. This facilitates indirect prompt injection, as a user could provide malicious instructions that the agent then codifies as a permanent repo-wide rule.
- Ingestion points: Untrusted content is ingested from user prompts via the SKILL.md workflow triggered by phrases like 'remember', 'always', or 'never'.
- Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when writing to AGENTS.md or project-context/ files.
- Capability inventory: The skill uses file-writing capabilities to modify AGENTS.md and various files within the project-context/ directory.
- Sanitization: There is no evidence of input validation, escaping, or sanitization to prevent the storage of instructions that might override system prompts or safety guidelines.
Audit Metadata