The Agent Skills Directory

[SAFE]: The skill is designed for structural planning and copywriting. It does not include any executable code, scripts, or automation that would interact with the host system's shell or network.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied offer briefs, documentation, and testimonials. While this is a common attack surface, the skill implements a robust source-policy.md and final-qa.md framework. These instructions mandate that every claim must trace directly to the provided offer map and explicitly forbid the agent from inventing content or following instructions not supported by the source facts.
Ingestion points: User-provided offer inputs, product documentation, research, and testimonials.
Boundary markers: The Entry Contract and Promo-Page Spine define strict boundaries and routing for out-of-scope tasks, preventing the agent from being diverted by malicious data.
Capability inventory: The skill is restricted to generating text (Markdown) and does not call any external tools or perform file-system modifications beyond reading context files.
Sanitization: The source-policy.md acts as a logical filter, requiring the agent to disregard unsupported claims or 'persuasion assets' not found in the source material.
[EXTERNAL_DOWNLOADS]: The skill includes a static link to a Telegram channel for user support. This is a visible, plain-text reference to a well-known service and does not trigger any automated downloads or background network activity.

a1-promo-page