The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted binary data from .cap files (Converted Applet Package), which presents a potential surface for indirect prompt injection if the ingested content contains malicious instructions intended to manipulate the agent's output.
Ingestion points: The skill reads external .cap files sourced from hardware dumps, firmware images, or card extractions as described in the "CAP 来源与文件识别" section of SKILL.md.
Boundary markers: The skill describes validation of binary structures and magic numbers (e.g., 0xDECAFFED), but does not define explicit delimiters or instructions to ignore embedded natural language commands that might appear in debug components or string tables.
Capability inventory: The skill utilizes system tools via subprocess execution including git, python3, java, unzip, and systemctl for environment setup and analysis.
Sanitization: The instructions focus on structural tag and size validation of the CAP components but do not prescribe sanitization for text-based metadata or debug strings within the files.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install several third-party security tools from public repositories and package registries.
Evidence: Instructs cloning tools from GitHub (e.g., martinpaljak/ant-javacard, benallard/caprunner, osmocom/pysim, frankmorgner/vsmartcard) and using standard package managers (e.g., pip install pyscard, apt install default-jdk).
[COMMAND_EXECUTION]: The skill involves the execution of various shell commands to perform analysis and manage system services for card interaction.
Evidence: Commands include java -jar gp.jar for card management, python3 readcap.py for parsing, and systemctl restart pcscd for managing the Smart Card resource manager.

re-javacard