The Agent Skills Directory

[SAFE]: The skill implements a rigorous requirement-gathering protocol that includes mandatory human confirmation and multiple validation gates. This structure prevents the agent from proceeding with ambiguous or potentially unsafe instructions.
[SAFE]: The skill includes an ethical 'grill' and evaluation process designed to detect and flag legal risks like deceptive advertising or dark patterns in user-provided briefs. This acts as a significant safety layer for the copywriting pipeline.
[SAFE]: Technical analysis shows no evidence of remote code execution, external data exfiltration, or credential exposure. All dependencies and agent references are contained within the local toolkit environment.
[SAFE]: Regarding indirect prompt injection, the skill processes user-supplied briefs but mitigates risk through structured data extraction and a mandatory completeness gate.
Ingestion points: Raw user briefs are processed in SKILL.md via the brainstorming and express protocols.
Boundary markers: Input is encapsulated within a structured 'Understanding Summary' artifact.
Capability inventory: The skill is limited to text manipulation and calls to other internal toolkit agents; it lacks network, file-write (beyond artifacts), or subprocess capabilities.
Sanitization: Content is validated against the intake-completeness-checklist.md and legal ethics standards.

copywriting-intake