The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (e.g., yt-dlp, node) using user-provided data (<youtube-url>) directly inside command arguments. This pattern is highly susceptible to shell command injection if the underlying platform does not strictly sanitize the input placeholders, potentially allowing arbitrary code execution on the host system.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8). It fetches and processes untrusted transcripts and subtitles from external sources (YouTube) without using boundary markers or sanitization logic to prevent embedded instructions from overriding the agent's behavior.
Ingestion points: Video transcripts and subtitles fetched from YouTube in Step 1 of the pipeline.
Boundary markers: None present; external data is processed directly as text for analysis.
Capability inventory: The skill possesses the capability to execute shell commands, perform network operations, and write to the local file system.
Sanitization: No sanitization or content filtering is performed on the ingested transcript data.
[EXTERNAL_DOWNLOADS]: The skill utilizes yt-dlp to fetch external content from YouTube and references local scripts (fetch-transcript-apify.js, fetch-transcript.js) for transcript retrieval. While YouTube is a well-known service, the method of fetching and processing this external data introduces significant attack surface.
[DATA_EXFILTRATION]: The skill has the capability to write data to local storage (YOUR_VAULT_PATH) and access temporary system directories ($TEMP). If compromised via a malicious transcript or command injection, these capabilities could be abused to exfiltrate or manipulate local files.

short-form