The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external partner data provided through an intake template or JSON file. This creates an entry point where untrusted data (such as partner names or descriptions) could contain instructions designed to influence the agent's evaluation or outputs.
Ingestion points: User-filled assets/partnership_intake_template.md and equivalent JSON files.
Boundary markers: Absent. The skill instructions do not specify the use of delimiters or directives for the agent to ignore instructions embedded within the partner data.
Capability inventory: Execution of analysis scripts including scripts/partner_tier_classifier.py, scripts/joint_gtm_planner.py, and scripts/revshare_modeler.py which generate text-based recommendations.
Sanitization: Absent. The scripts perform data parsing and logical classification but do not include safety filtering for prompt injection patterns.
[SAFE]: All scripts included in the skill utilize only the Python standard library and do not perform any network operations, sensitive file access, or dynamic code execution.

partnerships-architect