The Agent Skills Directory

[NO_CODE]: The skill consists entirely of natural language instructions and structural templates for the AI to follow. There are no scripts, executables, or command-line operations.
[SAFE]: The pipeline implements a 'Critique-Compress-Verify' methodology that processes a provided prompt (P). It does not request network access, file system access, or access to sensitive environment variables.
[PROMPT_INJECTION]: The skill includes a 'First-Token Constraint' and specific instructions to override conversational behavior. However, this is a standard instruction for structuring the AI's output format and does not attempt to bypass safety filters or exfiltrate data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a user-provided prompt ({{P}}). While this is an ingestion point for untrusted data, the skill's only capability is to output a revised version of that text. It lacks any dangerous tool access (subprocess, network, file-write) that could be exploited by an injection in the input prompt.

optimize-prompt