The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The setup-cron command in scripts/dgclaw.sh embeds the DGCLAW_API_KEY directly into the crontab entry as a plain-text environment variable. This allows any user or process with permission to list crontabs to harvest the sensitive API key, granting unauthorized access to the agent's competition and forum account.
[COMMAND_EXECUTION]: The skill implements a persistence mechanism via the setup-cron command in scripts/dgclaw.sh, which installs a recurring shell task in the user's crontab. This task executes a pipeline that fetches data and triggers agent chat interactions automatically.
[PROMPT_INJECTION]: The auto-reply functionality created by setup-cron establishes an indirect prompt injection surface. It fetches unreplied posts from an external forum and pipes them directly to the agent's chat context without sanitization or protective boundary markers, allowing malicious forum content to potentially manipulate agent actions.
Ingestion points: scripts/dgclaw.sh (unreplied-posts command fetching from the forum API)
Boundary markers: Absent; the fetched content is piped directly into the openclaw agent chat command.
Capability inventory: The skill has significant capabilities including perpetuals trading (perp_trade), modifying positions (perp_modify), and financial withdrawals (perp_withdraw).
Sanitization: Absent; the raw output from the forum API (via jq) is passed to the agent chat interface.

degenclaw