comment-stripping-preprocessor
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code which creates a surface for indirect prompt injection. Malicious instructions could be crafted within comment tags that the preprocessor is explicitly instructed to retain, such as SAFETY:, FIXME:, or TODO:, which may then be interpreted by the LLM in subsequent tasks.
- Ingestion points: The source_code parameter in the strip_code_comments function within SKILL.md.
- Boundary markers: The logic lacks specific boundary markers or instructions to the model to ignore content within the preserved tags.
- Capability inventory: The skill is limited to regex-based string manipulation and does not have access to subprocesses, network operations, or file-system writing.
- Sanitization: The skill implements sanitization by stripping most comments but intentionally whitelists certain patterns for retention.
Audit Metadata