The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill performs dynamic script generation and execution in scripts/extract_keys.py. It constructs a Python host script and a JavaScript instrumentation script from internal templates, writes the host script to a temporary file, and executes it to facilitate process hooking and key extraction.
[INDIRECT_PROMPT_INJECTION]: The skill processes WeChat chat logs, which contain untrusted data from third parties. If an agent analyzes these logs, there is a risk that malicious messages could act as indirect prompt injections to override the agent's instructions.
Ingestion points: Chat data is read from decrypted SQLite databases by scripts/vault_cli.py, scripts/export_chat.py, and scripts/wechat_digest.py.
Boundary markers: Absent. The skill does not use specific delimiters to isolate user-generated chat content from the rest of the prompt.
Capability inventory: The skill has access to shell command execution (codesign, pip, pgrep), process instrumentation via Frida, and local file system writes.
Sanitization: Absent. No filtering or escaping is applied to the chat message content before it is presented to the agent.
[COMMAND_EXECUTION]: scripts/extract_keys.py uses subprocess.run to execute system commands, including codesign for ad-hoc signing of application binaries and pgrep for process discovery. These are used to bypass system library validation and attach debuggers to the local WeChat process.
[EXTERNAL_DOWNLOADS]: The skill uses pip to automatically install the frida, frida-tools, and pycryptodome packages from the Python Package Index (PyPI) if they are missing at runtime.

yichen-wechat-local-vault