The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to read sensitive API credentials (API key, secret, and access tokens) from a plain-text JSON file located at ~/.claude/skills/x-post/x.key. While this is common for CLI tools, it exposes credentials to any process that can read the user's home directory.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the xdk Python package. This is an unverifiable dependency as the primary package with this name on public registries is unrelated to the X (Twitter) API, which may lead to installation failures or unexpected behavior.
[DATA_EXFILTRATION]: The skill transmits user-provided text and local media files to external endpoints at upload.twitter.com and api.x.com. This is the core purpose of the skill and is documented for transparency.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection where malicious instructions in processed data could attempt to trigger unauthorized posts.
Ingestion points: The text argument in x-post.py (passed via Bash) accepts arbitrary string data that the agent may pull from external sources.
Boundary markers: The SKILL.md file contains a 'Rules' section that mitigates this risk by explicitly instructing the agent to show the final text to the user and obtain confirmation before posting.
Capability inventory: The script has the capability to upload media and create posts via the X API in x-post.py.
Sanitization: No programmatic sanitization or filtering of the input text is performed in the script; safety relies entirely on the agent following the confirmation protocol.

x-post