awesome-openclaw-skills / idea

Skill instructs agent to execute `~/clawd/scripts/explore-idea.sh` with user-supplied idea description (line 81). The script is not bundled with SKILL.md and its contents are unknown. Without visibility into what explore-idea.sh does, the agent could be executing arbitrary commands. The skill also references `--dangerously-skip-permissions` flag (line 54), indicating the script intentionally bypasses safety controls. No allowed-tools declared, but the instruction directs concrete system action (shell script execution) that serves a purpose misaligned with typical agent capabilities — launching background sessions, managing tmux, sending files to Telegram, and queuing notifications are all operations that require explicit user awareness and approval.

criticalskills/idea:L49-55,81

Line 53 states the script 'Unsets OAuth env vars to use Claude Max'. Unsetting authentication tokens or credentials is a red flag for either credential theft or privilege escalation. The skill directs this action without explaining why OAuth vars must be removed or what 'use Claude Max' means in this context. This could enable unauthorized API access or token harvesting.

highskills/idea:L53

Line 54 explicitly references `--dangerously-skip-permissions` flag passed to claude CLI. This indicates the explore-idea.sh script intentionally disables safety controls. The skill body directs the agent to invoke this script, effectively instructing the agent to run code with safety mechanisms disabled. This is a direct attempt to weaken existing security controls.

highskills/idea:L54