style-review
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local CLI utility,
agent-style, to perform deterministic audits (--audit-only) and document comparisons (--compare). - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user Markdown files (
FILE) which are then processed by the LLM for semantic review and revision. As seen inreferences/revision-prompt.md, the skill uses---delimiters to isolate the untrusted content. To prevent the source text from overriding instructions, the system prompt enforces "hard invariants" that prohibit the model from adding new claims or connectives not present in the original text. The skill has the capability to write the revised content to a new file and execute the audit tool on the result. - [EXTERNAL_DOWNLOADS]: References the vendor's repository at
github.com/yzhao062/anywhere-agentswithin documentation and test fixtures (references/fixture-prose/regression-w3-rule-g.md) for verifying link and heading formatting.
Audit Metadata