jargon-extractor
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documentation files from the repository to extract jargon, creating a potential attack surface for indirect prompt injection.
- Ingestion points: Documentation and project management files located in directories such as docs/, doc/, documents/, and decision-record folders (adr/, decisions/, etc.) as specified in SKILL.md.
- Boundary markers: The subagent task templates in SKILL.md lack explicit boundary markers or 'ignore instructions' warnings when passing file content to extraction subagents.
- Capability inventory: The skill has permissions to write to the file system (JARGON.md and agent behavior files) and execute local shell commands and scripts.
- Sanitization: There is no explicit sanitization or filtering performed on the content extracted from documents before it is used to define terms in the glossary.
- [COMMAND_EXECUTION]: The skill performs standard local shell operations to manage its environment and update documentation.
- Evidence: The skill uses mktemp to create temporary scratch directories and executes a bundled Python script (scripts/merge_jargon.py) to handle the merging of glossary entries on disk.
Audit Metadata