okf-bundle
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from Markdown frontmatter across multiple files and aggregates it into a single index.md or table. Malicious instructions embedded in a file's frontmatter (e.g., in a title or description) could be propagated to the summary files, potentially influencing the behavior of AI agents that process those summaries later.
- Ingestion points: The skill reads all .md files in the specified --bundle-root directory using Path.read_text in scripts/okf_index.py and scripts/okf_validate.py.
- Boundary markers: For table injection, the skill uses explicit markers and to delimit generated content, which helps distinguish it from manually written text.
- Capability inventory: The skill has file writing capabilities through Path.write_text in scripts/okf_index.py to create or update index files and tables.
- Sanitization: The _md_escape function in scripts/okf_index.py escapes pipe characters (|) to maintain Markdown table structure, but there is no broader sanitization to detect or neutralize natural language instructions intended for prompt injection.
Audit Metadata