xy-human-touch
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided text for 'AI flavor' detection.
- Ingestion points: The agent receives arbitrary text from the user for analysis in Phase 1 and Phase 2 of
SKILL.md. - Boundary markers: The reporting structure in Phase 3 uses markdown block quotes to delimit original user sentences.
- Capability inventory: The skill can execute a local search script (
scripts/atoms-search.py) and access local files (references/atoms.jsonl). - Sanitization: The instructions guide the agent to perform specific linguistic analysis and use a structured reporting format, mitigating accidental instruction following.
- [COMMAND_EXECUTION]: The skill instructions in
SKILL.mdexplicitly direct the agent to execute a local Python script:python3 <本 skill 目录>/scripts/atoms-search.py. This script searches the skill's atomic knowledge base for evidence supporting its analysis. - [EXTERNAL_DOWNLOADS]: The
scripts/atoms-search.pyscript is configured to perform network requests tohttps://api.xyskill.xyz/v1/atoms/searchto fetch knowledge data. This domain is associated with the skill's author infrastructure and serves as a remote repository for the skill's data.
Audit Metadata