writing-clearly-and-concisely
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script (
scripts/check_edit.py) to ensure that prose modifications do not inadvertently change non-prose elements like code blocks, URLs, or metadata. The script utilizes standard libraries for text analysis and file reading. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process and rewrite user-provided text files, creating a surface for indirect prompt injection. Maliciously crafted input could theoretically attempt to influence the agent's behavior during the editing process.
- Ingestion points: The agent reads and analyzes local markdown files (e.g.,
original.md) provided by the user. - Boundary markers: The instructions explicitly define "frozen" sections (front matter, fenced code, shortcodes, HTML comments) and provide clear delimiters to help the agent distinguish between editable prose and structured data.
- Capability inventory: The skill has the capability to read local files and execute a specific local Python script using the
python3command. - Sanitization: The
check_edit.pyscript acts as a secondary verification layer, using regular expressions to validate that sensitive data structures remain byte-identical after the agent's edits.
Audit Metadata