simple-english
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted technical text (e.g., READMEs, documentation) provided by the user to perform rewriting tasks. This creates an attack surface for indirect prompt injection where malicious instructions embedded in the input text could attempt to override agent instructions. However, the risk is minimal because the skill lacks access to destructive tools or external networks.
- Ingestion points: Technical documentation, runbooks, and error messages provided by the user for rewriting or checking (SKILL.md).
- Boundary markers: The skill establishes an "Untouchables" policy (SKILL.md) which explicitly directs the agent to ignore code blocks, identifiers, and quoted error messages during the rewriting process.
- Capability inventory: The skill is limited to text analysis and natural language generation. No tool calls, shell executions, or network requests are defined.
- Sanitization: No input sanitization or explicit boundary markers (like delimiters) for user content are specified.
- [NO_CODE]: The skill contains no executable code, scripts, or binaries. It is composed entirely of natural language instructions in Markdown format and a simple YAML configuration file for the interface.
- [SAFE]: The skill does not contain any patterns associated with data exfiltration, credential harvesting, obfuscation, or persistence. All instructions are dedicated to the legitimate task of linguistic refinement and technical documentation standards.
Audit Metadata