ccf-skill-forger
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes several local Python scripts to automate skill management tasks. These include 'init_skill.py' for creating new skill structures, 'quick_validate.py' for syntax and structural checks, and 'check_path_privacy.py' for identifying sensitive local data. These operations are conducted within the local file system.
- [INDIRECT_PROMPT_INJECTION]: Because the skill audits and processes other skill files, it is exposed to potential injection attacks from the content it analyzes. However, it implements a structured audit workflow to mitigate these risks. (1) Ingestion points: Project skill files, including SKILL.md and referenced documentation. (2) Boundary markers: The skill uses predefined checklists and deterministic tools for validation rather than directly executing the analyzed logic. (3) Capability inventory: Local file system access and execution of trusted project scripts. (4) Sanitization: A mandatory check using 'check_path_privacy.py' is required to sanitize absolute paths and personal identifiers.
- [DYNAMIC_EXECUTION]: The skill instructions provide PowerShell logic to dynamically resolve tool paths at runtime based on environment variables like $CODEX_HOME. This dynamic resolution is standard for maintaining portability across different developer environments.
Audit Metadata