The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script chapters.py dynamically modifies the Python module search path at runtime. It uses sys.path.insert(0, ...) to include a directory relative to the script's own path (../yt-edit) and subsequently imports the deadair module. Loading executable code from computed relative paths is a risky practice as it assumes the parent directory structure is secure and cannot be manipulated by an attacker to inject malicious modules.
[COMMAND_EXECUTION]: The skill requires the agent to execute a local Python script using the python3 command. This script performs file system operations, specifically reading transcript files provided as command-line arguments.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data in the form of video transcripts (.srt, .vtt, or .json files). This represents a potential attack surface for indirect prompt injection if malicious instructions are embedded in the transcript text.
Ingestion points: The chapters.py script reads the contents of the file path passed to it in sys.argv[1] via the load() function.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to ignore potentially malicious embedded content within the transcript.
Capability inventory: The script has the capability to read files from the local filesystem and print formatted text to the agent's output buffer.
Sanitization: The script performs basic sanitization during keyword extraction using the regular expression [a-z']{4,}, which restricts the characters used to generate draft titles to lowercase letters and apostrophes, limiting the risk of control character injection into the final output.

yt-chapters