The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts arbitrary user text to generate infographic models, which are then passed to downstream tools. This creates an attack surface where malicious instructions embedded in the source text could attempt to influence the agent's behavior.
Ingestion points: User-supplied queries and information units processed in the Content model section of SKILL.md.
Boundary markers: The instructions include a 'Visual system' and 'Content model' that structure the output, along with an explicit constraint: 'Never include credentials, private identifiers, or unrelated context in the prompt.'
Capability inventory: The skill primarily models data for processon-prompt, processon-diagram, or processon-mindmap tools; it does not execute system commands, write files, or perform network operations directly.
Sanitization: The workflow requires extracting specific metrics and facts while removing 'repeated prose and decorative filler,' which provides a form of structural filtering.
[SAFE]: The skill demonstrates high security maturity by including explicit sections for 'Accuracy and boundaries', 'Rules' (Privacy security, Least privilege), 'Gotchas' (Version drift, Evidence expiration), and a comprehensive 'Validation checklist' to prevent unauthorized actions and data leakage.

processon-infographic