The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and reasoning over external user specifications and requirements documents (PRDs), which presents an attack surface for indirect prompt injection.
Ingestion points: The skill reads user-provided specifications and repository content during the 'Understand' and 'Intake' phases, specifically managed by instructions in SKILL.md and references/_shared/specification-pressure.md.
Boundary markers: The skill does not define specific string delimiters for user input in its core prompt instructions, relying instead on the agent's logical classification of statements.
Capability inventory: The agent has capabilities for code generation ('Build Mode'), system auditing ('Audit Mode'), and observing rendered output when a runnable environment is provided.
Sanitization: The instructions in specification-pressure.md provide a strong defense by requiring the agent to categorize statements (e.g., Hard Rule, Recommendation, Example) and prohibiting the automatic promotion of suggestions to requirements, which effectively mitigates the impact of instructions embedded within processed data.
[EXTERNAL_DOWNLOADS]: The documentation includes references to external code repositories used as design inspirations.
Evidence: references/_shared/third-party-notices.md contains links to github.com/Nutlope/hallmark and github.com/nextlevelbuilder/ui-ux-pro-max-skill.
Context: These links are provided for legal attribution and provenance of adapted design concepts; the skill does not include scripts that automatically download or execute content from these locations.

use-all-skills