The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill implements robust privacy protections to prevent accidental data leaks. The scripts/assess_profile.py script uses regular expressions to detect and block sensitive information (such as email addresses, phone numbers, IP addresses, secrets, and local file paths) from being included in public profiles. It also uses restrictive file permissions (0o600) for private data storage.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, specifically the Moltbook social network and GitHub issues. It mitigates injection risks by treating all external content strictly as data, using rigid JSON schema validation, and explicitly instructing the agent to ignore any commands or instructions embedded in candidate profiles or posts.
[COMMAND_EXECUTION]: The skill uses local Python scripts for assessment, validation, and matching. It employs a hash-bound approval mechanism for all public actions, requiring the agent to show the owner the exact content and target, and obtaining consent for the specific SHA-256 hash of that action before execution.
[EXTERNAL_DOWNLOADS]: The skill communicates with the Moltbook API and GitHub API to exchange profiles. These network operations are scoped to specific endpoints and require explicit owner authorization or the use of owner-managed credentials. It also includes checks to prevent communication with non-loopback proxies.
[DYNAMIC_EXECUTION]: Internal modules are loaded using importlib.util to maintain code modularity. This is used solely to reference sibling scripts within the skill's own directory and does not involve the execution of arbitrary or untrusted external code.

find-complementary-founders