The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages sensitive session data stored in 'cookies/tk_uploader/account.json'. Accessing and relying on session cookie files is a sensitive operation, as these files contain credentials that could be used to hijack a user's TikTok account. Although the instructions explicitly forbid reading or exfiltrating the file, the skill's reliance on this path marks a sensitive data boundary.
[COMMAND_EXECUTION]: The skill requires the agent to execute Python code blocks that call into local repository scripts, specifically 'uploader/tk_uploader/main_chrome.py'. This execution is necessary for the skill's primary function but grants the agent the ability to run arbitrary logic defined within the repository's files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of video titles and tags provided by the user or extracted from filenames. This creates a vulnerability surface where malicious content in a title or tag could influence the behavior of the browser automation tool (Patchright/Playwright) or the underlying platform during the upload process.
Ingestion points: User-provided video titles, tags, and file paths (referenced in SKILL.md).
Boundary markers: None identified; user input is passed directly as arguments to the TiktokVideo constructor.
Capability inventory: The skill executes local Python scripts, reads local session files, and performs network operations via browser automation.
Sanitization: No sanitization or escaping mechanisms are described in the instructions to handle special characters or potential injection payloads in the metadata.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic execution by instructing the agent to run Python code that imports modules and classes at runtime to interact with the TikTok web interface. The use of 'Patchright' (a stealth-focused fork of Playwright) indicates an intent to automate interactions while bypassing platform bot-detection mechanisms.

tiktok-upload