The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill utilizes directives such as 'without hesitation or deterrence,' 'overcome blockers,' and 'no excuses' to force the agent to bypass standard safety and reasoning protocols. This behavioral override is designed to suppress the agent's internal analysis and safety checks.
[COMMAND_EXECUTION] (HIGH): The combination of the Bash tool and instructions to 'Find another way' when blocked strongly encourages the agent to perform unsafe operations, such as privilege escalation or disabling security controls, if they interfere with rapid execution.
[INDIRECT_PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to instructions embedded in processed data (e.g., malicious scripts or task descriptions) because it is explicitly told to prioritize speed over research and analysis. Ingestion points: Processes untrusted task descriptions and codebases via Read and Bash tools. Boundary markers: Absent; the skill discourages the use of strategic planning or analysis. Capability inventory: Bash, Write, Edit, Read (High write/execute capability). Sanitization: None; the '15-Minute Rule' and 'Good Enough Test' prioritize immediate shipping over validation.

swift-executor