wait-what
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions define a communication style (ASD-STE100) and reference terminology from local glossary files. No malicious patterns, bypasses, or injection attempts were detected in the instructions. The frontmatter includes 'disable-model-invocation: true', which is a security-positive configuration that restricts the model from calling tools while this skill is active.
- [NO_CODE]: The skill consists entirely of markdown and configuration files. There are no scripts (Python, Node.js, Shell), binaries, or external dependencies to execute.
- [INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to process content from 'GLOSSARY.md' or 'GLOSSARY-MAP.md'. While this represents a surface for indirect prompt injection from repository data, the skill lacks any exploitable capabilities such as file system writes, tool access, or network operations, effectively mitigating the risk.
- Ingestion points: 'GLOSSARY.md', 'GLOSSARY-MAP.md' (referenced in SKILL.md).
- Boundary markers: None present in the prompt instructions to delimit glossary content.
- Capability inventory: None; the skill provides no scripts, tools, or binary executables.
- Sanitization: None present.
Audit Metadata