The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates an interactive setup that ingests user-provided outcomes, proofs, and voice settings, which are then stored in brand-config.json and SOUL.md. This content is intended to be consumed by downstream skills (evp-craft, evp-brief), creating a potential vector for indirect prompt injection if the user input contains malicious instructions designed to influence the behavior of those downstream tools.
Ingestion points: User input collected during Steps 3 through 7 of the onboarding workflow (SKILL.md).
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions for the data being persisted to the configuration files.
Capability inventory: The skill uses the Write tool to save data to the filesystem and the Read and Grep tools to check for existing configurations (SKILL.md).
Sanitization: No explicit validation, escaping, or filtering of the user-provided text is mentioned prior to the file-write operation.
[EXTERNAL_DOWNLOADS]: The skill instructions specify a dependency on cmj-hub/claude-psp. This is a vendor-owned resource from the same author as the skill, documented as a necessary prerequisite for accurate EVP (Employee Value Proposition) generation.

evp-onboarding