teach-pack
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from multiple sources to facilitate lessons, creating a surface for indirect prompt injection.
- Ingestion points: Learner answers (chat input), project source code, and historical artifacts (commits, PR comments) are read into the agent's context for analysis and 'grading'.
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when handling external content.
- Capability inventory: The skill utilizes file-writing capabilities to maintain a learning record in
.skills/teach-pack/and command execution to run an 'oracle' (e.g., project-specific linters or scripts) to verify learner predictions. - Sanitization: There is no evidence of input validation or sanitization for the content processed from project files or user responses before interpolation into prompts.
Audit Metadata