The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown-based instructions for the agent. It does not include any executable scripts, shell commands, or platform-specific tool definitions.
[SAFE]: The instructions promote security-positive behavior, specifically advising the agent not to expose credentials, not to bypass restrictions using other tools, and to seek explicit user approval for destructive, costly, or out-of-scope actions.
[INDIRECT_PROMPT_INJECTION]: The skill describes a workflow for processing untrusted data (user requests and project policies). While it lacks explicit delimiters or sanitization instructions for this data, there are no capabilities or tools associated with this skill that could be exploited. This represents a benign instructional surface.
[DATA_EXPOSURE]: The skill includes explicit instructions to protect secrets, stating that the agent should refer to secrets through secure mechanisms rather than exposing them in commands.

approval-gate