The Agent Skills Directory

[COMMAND_EXECUTION]: The setup.md file contains a shell script used to initialize the local storage directory at ~/discover/. The script correctly implements the principle of least privilege by applying restrictive permissions (chmod 700 for directories and chmod 600 for files) to protect the stored discovery data.
[DATA_EXFILTRATION]: The skill performs search queries against Google, Google News, and Reddit to find information related to user-defined topics. This is a documented core functionality, and the skill specifies that only topic keywords and query variants are sent externally.
[PROMPT_INJECTION]: As the skill ingests content from external web sources (Google, Reddit), it is theoretically susceptible to indirect prompt injection. However, the skill mitigates this risk through a 'Novelty Test' logic that filters incoming data for relevance and factual delta, and it does not allow the ingested data to trigger automated actions or code execution.
[SAFE]: The skill includes explicit instructions for the agent to seek user approval before modifying workspace configuration files (AGENTS.md, HEARTBEAT.md) or activating recurring discovery cycles.

Discover