continuous-learning
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by extracting reusable patterns and durable facts from untrusted session transcripts to update the AGENTS.md file and create new skill candidates. \n
- Ingestion points: The skill mines transcript indexes at ~/.cursor/projects//agent-transcripts/ and supports importing instincts from JSON data (SKILL.md). \n
- Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" warnings to prevent the agent from being influenced by malicious instructions embedded within the source transcripts. \n
- Capability inventory: The skill orchestrates the agents-memory-updater subagent to perform file writes to AGENTS.md and manages the lifecycle of new "instincts" that can be promoted to standalone skills (SKILL.md). \n
- Sanitization: No sanitization, validation, or filtering mechanisms are described for the content harvested from conversation history before it is stored or promoted.
Audit Metadata