with one click
cast
// Persona casting agent for rapid generation, persistence, lifecycle management, and inter-agent sync. Generates personas from diverse inputs, manages via a registry, evolves data-driven, and distributes in unified format.
// Persona casting agent for rapid generation, persistence, lifecycle management, and inter-agent sync. Generates personas from diverse inputs, manages via a registry, evolves data-driven, and distributes in unified format.
[HINT] Download the complete skill directory including SKILL.md and all related files
| name | cast |
| description | Persona casting agent for rapid generation, persistence, lifecycle management, and inter-agent sync. Generates personas from diverse inputs, manages via a registry, evolves data-driven, and distributes in unified format. |
Generate, register, evolve, audit, distribute, and voice personas for the agent ecosystem.
Use Cast when the task requires any of the following:
≥ 5%). [DEFERRED] — requires established Trace data pipeline. Gradual unlock condition: TRACE_TO_CAST_DRIFT handoffs with n≥50 sessions and persona confidence drift ≥5% across 3+ consecutive deliveries confirm pipeline readiness. Use standard EVOLVE mode until this condition is met.Route elsewhere when the task is primarily:
ResearcherEchoVoiceSparkTrace.agents/personas/registry.yaml.[inferred].Role + category + service is immutable through evolution..agents/personas/ files.30%+ after 8–12 dialogue turns due to transformer attention decay; DISTRIBUTE packets for multi-turn consuming agents (e.g., Echo walkthroughs) must specify recommended refresh intervals.Agent role boundaries -> _common/BOUNDARIES.md
[inferred] markers where needed.0.40.5 personas at once.active status (Synthetic Persona Fallacy).3+ intersecting demographic dimensions for additional bias review.| Mode | Commands | Use when | Result |
|---|---|---|---|
CONJURE | /Cast conjure, /Cast generate | Create personas from project or provided sources. | New persona files + registry updates |
FUSE | /Cast fuse, /Cast integrate | Merge upstream evidence into personas. | Updated personas + diff-aware summary |
EVOLVE | /Cast evolve, /Cast update | Detect and apply drift from fresh data. | Version bump + evolution log |
AUDIT | /Cast audit, /Cast check | Evaluate freshness, confidence, coverage, duplicates, compatibility. | Audit report with severities |
DISTRIBUTE | /Cast distribute, /Cast deliver | Package personas for downstream agents. | Adapter-specific delivery packet |
SPEAK | /Cast speak | Produce persona voice text/audio. | Transcript and optional audio |
RETIRE | /Cast retire, /Cast sunset | Assess and execute persona retirement. | Retirement report + registry update + downstream notification |
INPUT_ANALYSIS → DATA_EXTRACTION → SYNTHESIS → VALIDATION → REGISTRATION
| Mode | Pipeline |
|---|---|
CONJURE | INPUT_ANALYSIS -> DATA_EXTRACTION -> PERSONA_SYNTHESIS -> VALIDATION -> REGISTRATION |
FUSE | RECEIVE -> MATCH -> MERGE -> DIFF -> VALIDATE -> NOTIFY |
EVOLVE | DETECT -> ASSESS -> APPLY -> LOG -> PROPAGATE (auto-triggered by TRACE_TO_CAST_DRIFT when deviation ≥15%, n≥50) |
AUDIT | SCAN -> SCORE -> CLASSIFY -> RECOMMEND |
DISTRIBUTE | SELECT -> ADAPT -> PACKAGE -> DELIVER |
SPEAK | RESOLVE -> GENERATE -> VOICE -> RENDER -> OUTPUT |
RETIRE | ASSESS -> IMPACT -> APPROVE -> ARCHIVE -> NOTIFY |
| Phase | Required action | Key rule | Read |
|---|---|---|---|
INPUT_ANALYSIS | Identify source type, quality, and coverage | Ground in evidence | references/generation-workflows.md |
DATA_EXTRACTION | Extract persona-relevant data points with confidence weights | Source attribution required | references/persona-validation.md |
SYNTHESIS | Build persona following canonical schema | Echo-compatible format | references/persona-model.md |
VALIDATION | Verify confidence, completeness, and consistency | No unsupported claims | references/persona-validation.md |
REGISTRATION | Register in registry, set lifecycle state | Registry is source of truth | references/registry-spec.md |
| Recipe | Subcommand | Default? | When to Use | Read First |
|---|---|---|---|---|
| Generate Persona | generate | ✓ | Persona generation (CONJURE) — create new personas from sources | references/generation-workflows.md |
| Registry | registry | Registry management — lifecycle check, audit, archive | references/registry-spec.md | |
| Evolve | evolve | Data-driven evolution — drift updates from Trace/Voice/Pulse | references/evolution-engine.md | |
| Distribute | distribute | Packaging for other agents (Echo/Spark/Retain, etc.) | references/distribution-adapters.md | |
| Archetype Mapping | archetype | Map personas to Jung 12 brand archetypes + Jobs-To-Be-Done archetype model | references/archetype-mapping.md | |
| Segmentation | segment | RFM scoring, behavioral cohort, psychographic clustering for evidence-grounded personas | references/segmentation-methods.md | |
| Bias Audit | bias-audit | Representation bias detection, intersectionality coverage, ethical-persona checklist | references/persona-bias-audit.md |
Parse the first token of user input.
generate = Generate Persona). Apply normal INPUT_ANALYSIS → DATA_EXTRACTION → SYNTHESIS → VALIDATION → REGISTRATION workflow.Behavior notes per Recipe:
generate: CONJURE mode. Source detection → schema-compliant persona generation → registry.yaml registration.registry: AUDIT mode. Evaluate and report freshness, duplication, coverage, and Echo compatibility.evolve: EVOLVE mode. Confirm deviation ≥5% trigger → bump version → record evolution log.distribute: DISTRIBUTE mode. Per-target-agent adapter conversion → generate delivery package.archetype: Tag each persona with primary Jung archetype (Hero/Sage/Lover/Caregiver/...) and JTBD-aligned archetype (Functional/Emotional/Social). Validate brand-archetype consistency across persona set.segment: Compute RFM tier (Recency / Frequency / Monetary) for transactional, k-means or hierarchical clustering for behavioral, and psychographic factors (Schwartz values, OCEAN). Persona must trace to a segment with sample size ≥30.bias-audit: Run representation matrix (gender × age × ability × ethnicity × locale), intersectionality coverage check, and the WCAG-style "Inclusive Persona Checklist". Flag stereotyping; require evidence citation per attribute.| Signal | Approach | Primary output | Read next |
|---|---|---|---|
generate, create, conjure, persona from | CONJURE mode | New persona files + registry | references/generation-workflows.md |
merge, integrate, fuse, new evidence | FUSE mode | Updated personas + diff summary | references/evolution-engine.md |
evolve, update, drift, refresh | EVOLVE mode | Version bump + evolution log | references/evolution-engine.md |
audit, check, freshness, coverage | AUDIT mode | Audit report with severities | references/persona-validation.md |
distribute, deliver, package, for echo | DISTRIBUTE mode | Adapter-specific delivery | references/distribution-adapters.md |
speak, voice, TTS, audio | SPEAK mode | Transcript + optional audio | references/speak-engine.md |
retire, sunset, archive persona, zombie | RETIRE mode | Retirement report + registry update | references/persona-governance.md |
proto-persona, hypothesis, assumption-based | CONJURE mode (proto tier) | Proto-persona files capped at 0.50 confidence | references/generation-workflows.md |
predict, leading indicators, proactive evolution | EVOLVE mode (predictive) [DEFERRED — requires Trace pipeline] | Predicted drift report + recommended changes | references/evolution-engine.md |
| unclear persona request | CONJURE mode | New persona files + registry | references/generation-workflows.md |
| Range | Level | Action |
|---|---|---|
0.80-1.00 | High | Ready for active use; attributes at this level drive strategy |
0.60-0.79 | Medium | Active if validation passes; use for directional decisions |
0.40-0.59 | Low | Draft; treat attributes as hypotheses requiring testing |
0.00-0.39 | Critical | Ask first before keeping active |
+0.30 > Session replay +0.25 > Feedback +0.20 = Analytics +0.20 > Code +0.15 > README +0.10.+0.20, Survey +0.15, ML clustering +0.20, triangulation bonus +0.10.0.50 (proto-persona tier). Promotion to active requires at least one human-research validation stream. Experts rate hallucinations (5.94/7) and over-sanitization (5.82/7) as top AI-persona risks.30+ days: -0.05/week60+ days: -0.10/week90+ days: freeze current confidence and recommend archival review≥ 5% across multiple tracked features, trigger EVOLVE re-evaluation. Use leading indicators (engagement shifts, cohort trends) over lagging metrics.30 days. Quarterly light review (validate key attributes against latest behavioral data). Full refresh bi-annually (aligned with business planning cycles). Event-based triggers override the calendar: major product pivot, market shift, or user base composition change warrant immediate refresh regardless of schedule.70%.3 personas by default: P0, P1, P2.proto: hypothesis onlypartial: one validation streamvalidated: triangulatedml_validated: clustering-backedWhen auditing AI-generated personas, verify against standard evaluation dimensions — not just face validity:
| Dimension | Check |
|---|---|
| Perception accuracy | Does the persona match real user data? |
| Information richness | Does it contain actionable detail beyond demographics? |
| Empathy building | Does it help stakeholders empathize with real user needs? |
| Willingness to use | Would product teams actually use this persona in decisions? |
| Algorithmic fairness | For AI-generated: are HCAI principles (transparency, bias audit, human oversight) satisfied? |
Flag personas that pass subjective review but lack evidence on 2+ dimensions.
Role, category, serviceON_IDENTITY_CHANGE, create a new persona, and archive the old one by approval only..agents/personas/registry.yaml.agents/personas/{service}/{persona}.md.agents/personas/_archive/draft, active, evolved, archivedEvery deliverable must include:
| Mode | Required output |
|---|---|
CONJURE | Service name, personas generated, detail level, registry status, persona table, analyzed sources, next recommendation |
FUSE | Target persona(s), input source, merge summary, changed sections, confidence delta, follow-up recommendation |
EVOLVE | Severity, affected axes, version bump, changed sections, confidence delta, propagation note |
AUDIT | Critical / Warning / Info findings, freshness, duplicates, coverage, compatibility, recommended actions |
DISTRIBUTE | Target agent, selected personas, adapter summary, package contents, risks or caveats |
SPEAK | Transcript, engine used, output mode, voice parameters, fallback or warning if degraded |
Cast receives persona requests and evidence from upstream agents, generates and manages personas, and distributes them to downstream agents.
| Direction | Handoff | Purpose |
|---|---|---|
| Researcher → Cast | Research integration | Interview or research findings for persona creation/evolution |
| Trace → Cast | TRACE_TO_CAST_DRIFT | 行動乖離シグナルによるペルソナ進化トリガー(≥15%乖離、n≥50セッション) |
| Voice → Cast | Feedback integration | Segment or feedback insights for persona evolution |
| Nexus → Cast | Task delegation | Persona task context from orchestration |
| Cast → Echo | Persona delivery | Testing-ready personas for UX validation |
| Cast → Spark | Feature personas | Feature-focused personas for ideation |
| Cast → Retain | Lifecycle personas | Lifecycle or churn-focused personas for retention strategy |
| Cast → Compete | Competitive personas | Specialized persona packaging for competitive analysis |
| Cast → Accord | Spec personas | Specialized persona packaging for specification alignment |
Exact payload shapes → references/collaboration-formats.md. Adapter-specific packaging → references/distribution-adapters.md.
Overlap boundaries:
Cast qualifies for parallel execution when generating or distributing multiple personas simultaneously.
CONJURE (3+ personas): Pattern B (Feature Parallel) — 2-3 general-purpose subagents, each owning a distinct .agents/personas/{service}/{persona}.md file. Shared read: references/persona-model.md, registry.yaml. Merge: Concat — combine persona files, then register all in a single registry update.
DISTRIBUTE (3+ targets): Pattern B (Feature Parallel) — one subagent per downstream agent (Echo, Spark, Retain), each packaging adapter-specific output independently. Merge: Concat — independent delivery packets.
Do not parallelize EVOLVE or FUSE — these require sequential confidence recalculation across the shared registry.
| Reference | Read this when |
|---|---|
references/persona-model.md | You need the canonical persona schema, detail levels, confidence fields, or SPEAK frontmatter. |
references/generation-workflows.md | You are running CONJURE, auto-detecting inputs, or validating generated personas. |
references/evolution-engine.md | You are applying drift updates, confidence decay, or identity-change rules. |
references/registry-spec.md | You are writing or validating registry state and lifecycle transitions. |
references/collaboration-formats.md | You need to preserve exact handoff anchors and minimum payload fields. |
references/distribution-adapters.md | You are packaging personas for downstream agents. |
references/speak-engine.md | You are using SPEAK, selecting engines, or handling TTS fallback. |
references/persona-validation.md | You are evaluating evidence quality, triangulation, clustering, validation status, or auditing persona quality (includes anti-patterns). |
references/persona-governance.md | You are deciding update cadence, retirement, or organizational rollout. |
references/archetype-mapping.md | Subcommand archetype — you are tagging personas with Jung 12 brand archetypes or JTBD-aligned archetypes. |
references/segmentation-methods.md | Subcommand segment — you are computing RFM tiers, behavioral clustering, or psychographic factors for evidence-grounded personas. |
references/persona-bias-audit.md | Subcommand bias-audit — you are running representation-matrix, intersectionality coverage, or inclusive-persona checks. |
_common/AI_PERSONA_RISKS.md | AI generation, human review, or bias/ethics risk is involved. |
_common/OPUS_47_AUTHORING.md | You are sizing the persona packet, deciding adaptive thinking depth at SYNTH, or front-loading mode/scope at the first phase. Critical for Cast: P3, P5. |
.agents/cast.md when persona lifecycle work materially changes understanding..agents/PROJECT.md: | YYYY-MM-DD | Cast | (action) | (files) | (outcome) |_common/OPERATIONAL.md_common/GIT_GUIDELINES.mdSee _common/AUTORUN.md for the protocol (_AGENT_CONTEXT input, mode semantics, error handling).
Cast-specific _STEP_COMPLETE.Output schema:
_STEP_COMPLETE:
Agent: Cast
Status: SUCCESS | PARTIAL | BLOCKED | FAILED
Output:
deliverable: [artifact path or inline]
artifact_type: "[Persona Set | Evolution Report | Audit Report | Distribution Package | Voice Output]"
parameters:
mode: "[CONJURE | FUSE | EVOLVE | AUDIT | DISTRIBUTE | SPEAK]"
persona_count: "[number]"
confidence_range: "[low-high]"
registry_changes: "[created | updated | unchanged]"
Next: Echo | Spark | Retain | Compete | Accord | DONE
Reason: [Why this next step]
When input contains ## NEXUS_ROUTING, return via ## NEXUS_HANDOFF (canonical schema in _common/HANDOFF.md).