persona-construction
$
npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/persona-constructionConstructs adversarial personas for red teaming scenarios
- Generates realistic adversary profiles for targeted attacks
- Uses subagent execution to maintain perspective separation
- Leverages persona type and domain inputs to shape character traits
- Returns structured profiles with attack patterns and trigger phrases
SKILL.md
.github/skills/persona-constructionView on GitHub ↗
--- name: persona-construction description: Build a detailed adversarial persona with background, motivation, expertise, blind spots, and preferred attack patterns. execution: subagent prompt: ./prompt.md input: persona_type (string), artifact_domain (string) used-by: [red-teaming] --- # Persona Construction Builds detailed adversary personas for roleplay-based attacks. ## Execution Subagent — spawned via subagent-spawning/spawn-agent. ## Why Subagent Persona construction requires creative character building without contamination from the artifact's perspective. The persona must feel authentic and internally consistent. ## Input - **persona_type**: Category of adversary (hostile-reviewer, competing-lab, funding-skeptic, domain-outsider, or custom) - **artifact_domain**: The domain of the artifact being attacked (for expertise calibration) ## Output - **persona**: Complete adversary profile (name, background, motivation, expertise, blind spots) - **attack_style**: How this persona typically attacks (preferred patterns) - **trigger_phrases**: What aspects of artifacts particularly provoke this persona