building-skills-from-patterns
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for the agent to 'promote' repeated interaction patterns and user corrections into persistent
SKILL.mdfiles within the.cursor/skills/directory. This automated creation of persistent instructions based on potentially untrusted or adversarial user input poses a risk of encoding malicious behaviors into the workspace environment. - Ingestion points: The agent monitors user requests, observed multi-step workflows, and user corrections ('policy' statements) from the chat history.
- Boundary markers: The skill does not define any boundary markers or instructions for the agent to distinguish between legitimate workflow patterns and adversarial attempts to inject persistent instructions.
- Capability inventory: The agent is granted the capability to write new markdown files to the local file system (specifically
.cursor/skills/<slug>/SKILL.md). - Sanitization: There are no instructions for sanitizing or validating the patterns before they are drafted into a new skill, other than a brief note to avoid 'secrets or machine-specific paths'.
Audit Metadata