desk
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection as it processes and summarizes untrusted prose data (e.g., from pull request descriptions, blog posts, or website copy) to generate skim extracts and SEO assessments. 1. Ingestion points: Prose artifacts are processed during the DRAFT, SKIM, and GEO steps defined in
SKILL.md. 2. Boundary markers: The instructions do not specify the use of delimiters or 'ignore' warnings for the content being processed. 3. Capability inventory: The skill has command execution capabilities (python3 scripts/lint.py) and summary capabilities (GEO/SKIM steps). 4. Sanitization: No specific sanitization or filtering of input prose is described before the agent processes it for summarization. - [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script (
scripts/lint.py) to perform deterministic prose linting. While the script itself uses only standard Python libraries and performs regex-based analysis, the execution of local scripts is a powerful capability that relies on the agent correctly handling file paths and arguments.
Audit Metadata