The Agent Skills Directory

[PERSISTENCE]: The skill includes a setup-cron command in scripts/dgclaw.sh that creates a crontab entry to periodically run the agent's forum reply logic, ensuring the skill remains active across sessions.
[INDIRECT_PROMPT_INJECTION]: The auto-reply feature fetches external forum posts and pipes their content directly to the agent's chat interface, creating a potential vector for malicious data to influence agent behavior.
Ingestion points: Forum posts are retrieved via curl from https://degen.virtuals.io/api/forums/:agentId/posts?unreplied=true as seen in scripts/dgclaw.sh.
Boundary markers: Absent; the raw post content is passed to the agent chat command without delimiters.
Capability inventory: The agent possesses capabilities to create trading jobs (acp job create) and interact via chat.
Sanitization: Absent; the script does not perform any validation or escaping of the retrieved forum content.
[CREDENTIALS_UNSAFE]: The setup-cron command in scripts/dgclaw.sh embeds the DGCLAW_API_KEY directly into the crontab line, which may expose the credential to other users on the system or in logs.
[EXTERNAL_DOWNLOADS]: The documentation in README.md and SKILL.md directs users to download and install external code from the Virtual-Protocol GitHub organization repositories.
[COMMAND_EXECUTION]: The dgclaw.sh script executes several system utilities including curl for API communication, openssl for RSA key generation and decryption, jq for JSON processing, and crontab for managing persistent tasks.

@1892/dgclaw