check-terms
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: Technical review of the skill's logic confirms it is a passive text analysis tool. It uses local shell scripts (
scripts/check-termsandscripts/terms-table) and reference files (references/terms.tsv,references/terms-caution.tsv) to identify linguistic patterns. There is no evidence of network access, credential harvesting, or unauthorized file system operations. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text provided by users or found in files to identify phrasing issues.
- Ingestion points: The
scripts/check-termsscript reads file paths or standard input provided at runtime. - Boundary markers: The script uses
awkto empty the content of markdown code blocks andsedto strip inline code and bracketed text (「...」) before running detection patterns. This prevents the linter from mistakenly analyzing and acting upon code or quoted examples. - Capability inventory: The skill's primary capability is executing local shell scripts to generate a report of suggested phrasing changes. It does not have write access to the files it checks within the script logic itself.
- Sanitization: Input is pre-processed to remove structural markers (code/quotes) that might contain instructions, ensuring the agent only evaluates the prose content for linguistic suggestions.
Audit Metadata