The Agent Skills Directory

[NO_CODE]: The skill consists entirely of markdown instructions (SKILL.md, methods.md) and configuration files (SOURCE-MANIFEST.json, openai.yaml). There are no scripts, binaries, or executable commands included in the package.
[SAFE]: Analysis of the instructions reveals no attempts at prompt injection, obfuscation, or persistence. The skill manages workflow state and delegates tasks to other installed skills without performing dangerous operations itself.
[DATA_EXPOSURE_EXFILTRATION]: No hardcoded credentials, sensitive file paths, or network communication patterns were detected. The skill's configuration explicitly disables direct model invocation, providing a restricted execution environment.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-provided task descriptions to select methods, it lacks the necessary capabilities (such as file writing or network access) to be exploited. It functions as a logic layer for selecting appropriate agent behaviors.

atlas-mode