The Agent Skills Directory

[NO_CODE]: The skill documentation refers to and instructs the execution of code_changelog_tracker.py and create_changelog.py. However, the actual source code for these Python scripts is missing from the skill package, making their behavior unverifiable.- [COMMAND_EXECUTION]: The skill contains instructions for executing several shell commands, including running the missing Python scripts, starting a local HTTP server (python3 -m http.server), and forcefully terminating processes using a combination of lsof and kill -9.- [DATA_EXFILTRATION]: By encouraging the use of python3 -m http.server to host the reviews folder, the skill risks exposing all logged code changes and project documentation to any device on the local network without authentication.- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting and logging code modifications, which creates a surface for indirect prompt injection where malicious instructions in the source code could influence the agent during review cycles.
Ingestion points: The skill reads contents of project files (e.g., main.py, auth.py) via the CodeChangeLogger to generate documentation.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the logged code as untrusted data or to ignore embedded instructions.
Capability inventory: The skill scripts are capable of writing to the local filesystem and initiating network-accessible services.
Sanitization: The documentation does not specify any sanitization, filtering, or escaping of the code content before it is written to the markdown logs.

code-changelog