The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes internal Python scripts to validate link configurations and asset cards.
Evidence: Files 00_eva-link_本地模块连接.md and 02_eva-link-doctor_Link健康检查.md contain commands such as python3 "<EVA_SHARED_ROOT>/scripts/eva_link_check.py" which use absolute paths to scripts located within the skill's distribution directory.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes external module instructions.
Ingestion points: Processes module.md and eva.link.json from user-defined directories (local-modules/) as described in SKILL.md.
Boundary markers: The skill outlines strict boundaries in SKILL.md and includes a validation mechanism in 02_eva-link-doctor_Link健康检查.md that checks for hidden or evasive instructions within modules.
Capability inventory: It has the capability to run local validation scripts and perform file system writes in the project directory.
Sanitization: Utilizes SHA256 fingerprinting (approved_sha256) in the registry and strict validation scripts to ensure module integrity and user consent.
[DYNAMIC_EXECUTION]: The skill dynamically generates instruction sets and configuration files for new local workflows.
Evidence: 01_eva-link-builder_自定义Link生成.md handles the creation of eva.link.json and module.md files at runtime based on user interactions.

eva-link