The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local project commands such as pnpm diagram:check --inventory and other validation tools defined in local configuration files (AGENTS.md or a 'diagram contract'). These operations are consistent with its purpose as a development and documentation tool and do not involve unauthorized privilege escalation or remote code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of lesson prose to generate diagrams. It mitigates the risk of instructions embedded in the prose (Indirect Prompt Injection) by enforcing a 'Warrant' and 'Diagram gates' policy, which strictly requires that every element in the output must be explicitly traced to a sentence in the lesson and prohibits any inferred or invented content.
[SAFE]: The skill demonstrates defensive programming practices, such as refusing to operate if a required project contract is missing and limiting the scope of its file modifications to a single lesson destination. The use of non-English keywords in the metadata is legitimate and does not constitute obfuscation.

diagram-generator