The Agent Skills Directory

[DATA_EXPOSURE]: The skill includes a dedicated script (scripts/source_deposit_plan.py) designed to inventory local project files. Crucially, this script incorporates a secret scanner that identifies hardcoded credentials (e.g., AWS keys, private keys, API tokens) to prevent their accidental inclusion in copyright materials. File operations are strictly local, and the skill's instructions explicitly mandate redaction before processing sensitive content.
[COMMAND_EXECUTION]: The workflow utilizes two provided Python scripts (source_deposit_plan.py and materials_gate.py) to automate administrative tasks like file inventorying and material validation. These scripts use standard Python libraries, avoid dynamic execution (no eval or exec), and operate on the local file system without network access.
[PROMPT_INJECTION]: The skill contains strong instructional constraints and 'Non-negotiable boundaries' to prevent the agent from providing legal advice or guaranteeing registration success. These instructions are consistent with the skill's operational purpose and do not attempt to bypass core safety filters.
[EXTERNAL_DOWNLOADS]: All external references in the documentation point to official government portals (e.g., China Copyright Protection Center, US Copyright Office) or established peer projects. No remote scripts or packages are downloaded or executed during the skill's operation.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external data in the form of project source code, it does so for the purpose of line counting and documentation drafting. The workflow is guarded by multiple mandatory user confirmation checkpoints, and the validation script (materials_gate.py) enforces schema compliance, effectively mitigating the risk of malicious instructions within processed files influencing the agent's behavior beyond the intended document generation scope.

light-software-copyright