The Agent Skills Directory

COMMAND_EXECUTION (MEDIUM): The README describes a Python wrapper (code_changelog_tracker.py) that manages shell processes for honkit build and honkit serve. The use of subprocess or similar execution methods to run these commands poses a risk of command injection if variables like project_name, file_path, or reason are not strictly sanitized before being passed to the shell.
EXTERNAL_DOWNLOADS (LOW): Setup requires installing the honkit package globally via npm. While HonKit is a legitimate documentation tool, it is an external dependency from a non-trusted source (per the defined list), which introduces standard supply chain risks.
INDIRECT_PROMPT_INJECTION (LOW): The skill ingests untrusted code snippets and user-provided 'reasons' for changes to generate documentation.
Ingestion points: log_file_creation and log_file_modification methods (File: README.txt)
Boundary markers: None mentioned or implemented in the described usage.
Capability inventory: Shell execution of build tools and local web server hosting (File: README.txt)
Sanitization: No sanitization or escaping of the ingested code content is mentioned.
PRIVILEGE_ESCALATION (LOW): The documentation suggests a global npm installation (npm install -g), which frequently requires elevated privileges (sudo) on Unix-like systems, increasing the impact of any malicious code within the package or its installation scripts.

code-changelog