The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from Twitter/X, creating a vulnerability to malicious instructions embedded in that content. 1. Ingestion points: The skill reads live content from X through commands such as superx x:post, superx x:replies, superx engage:posts, and superx replies:received (File: references/commands.md). 2. Boundary markers: The instructions lack explicit delimitation or specific warnings to the model to ignore instructions found within the ingested content. 3. Capability inventory: The agent has the ability to execute shell commands via the Bash tool, write files through datasets:export, and schedule or publish content to social media (File: SKILL.md). 4. Sanitization: No sanitization logic for external content is described in the prompt instructions.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the superx-cli package from the npm registry using npm install -g superx-cli (File: SKILL.md). This is a vendor-owned resource used to facilitate the skill's operations.
[CREDENTIALS_UNSAFE]: The skill manages SuperX API keys (SUPERX_API_KEY) and stores authentication credentials locally in ~/.superx/credentials.json (File: references/commands.md). It provides instructions for manual key entry and environment variable configuration for CI/automated environments.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via the superx CLI tool to perform data retrieval, analytics, and account management actions as part of its core workflows (File: SKILL.md).

superx