apply-guidelines
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to download guidelines from an external source and apply them as authoritative rules, which establishes an attack surface. If the remote repository is compromised, the agent would execute the updated instructions. • Ingestion points: Fetches markdown documents from github.com/fang2hou/ai-coding-guidelines (SKILL.md, Step 1). • Boundary markers: Absent. The agent is explicitly told to treat the fetched content as conclusions and apply them as written. • Capability inventory: Shell access for git and mise commands, and file system write access during the remediation phase. • Sanitization: None.
- [EXTERNAL_DOWNLOADS]: Fetches the AI Coding Guidelines repository from GitHub. • Evidence: git clone --depth 1 https://github.com/fang2hou/ai-coding-guidelines in SKILL.md.
- [COMMAND_EXECUTION]: Executes shell commands to maintain the guidelines cache and run project-defined quality gates. • Evidence: Uses git for repository management and mise run check for project validation in SKILL.md.
Audit Metadata