Add `find-perspective` skill — guided intellectual-signature excavation

[SnowboardTechie]

Problem / Motivation

Athena Notes has no skill that helps a user discover and articulate their intellectual signature — the thinkers, ideas, and experiences that shape their worldview, the mission that ties them together, and the 3–5 core ideas they could build a body of work around.

Closest analogs are partial:

• prism does pattern/contrast/naming refraction but produces no artifact and isn't phase-gated
• weekly-planning, session-review, and workday-planning touch on themes but never excavate worldview or produce a positioning blueprint
• The brand-voice plugin codifies existing voice but doesn't excavate from a blank slate

A user wanting this capability today has to paste an external prompt (e.g. Dan Koe's "Perspective Architect", Jan 2026) into a chat, with no integration with their vault, prior brand artifacts, or Athena's spoke-based research/synthesis/capture pipeline.

Proposed behavior

A new skill, find-perspective, that runs a guided multi-phase Q&A flow:

1. Excavation — surface the thinkers, ideas, and problems that shape the user's thinking.
2. Research expansion — research named thinkers via sage; surface core worldview + key concepts.
3. Synthesis & mission — refract patterns/contrasts via prism; iterate a mission statement until the user owns it.
4. Core idea crystallization — identify 3–5 "book-worthy" ideas with synthesis, articulation, and depth tests.
5. Blueprint write — scribe writes a permanent blueprint note to the vault.

The skill orchestrates four existing spokes (archivist for vault mining, sage for research, prism for synthesis, scribe for the blueprint write). It owns the Q&A flow, approval gates, and synthesis turns itself.

When the user has prior brand artifacts in their vault, the skill enters a refine branch (deepen / update existing) rather than re-excavating from scratch.

Scope

In scope

• New skill at plugins/athena-notes/skills/find-perspective/SKILL.md
• Phase 0 vault scan via archivist to detect prior brand/identity artifacts
• Phase 0 conversational prompt asking the user about active publication/streaming gates; surface in the blueprint's Next Steps
• Refine-vs-fresh branch based on Phase 0 findings
• Approval gates before mission lock-in (Phase 3) and before the vault write (Phase 5)
• Output: single Markdown note in the user's vault, structured with Mission / Stakes / Vision / Core Ideas (with synthesis tables) / Source Thinkers / Content Development Sequence / Anti-Vision / Next Steps
• Single SKILL.md with phased flow internally (matches weekly-planning, meeting-sync, issue-create, issue-work precedent)

Out of scope

• A new spoke. The work composes cleanly across four existing spokes, and AGENTS.md says interactive Q&A flows live in skills, not spokes
• Modifying or replacing the brand-voice plugin (different surface — that plugin operationalizes existing voice)
• Extending ~/.claude/athena/identity.md schema. Identity stays config; the blueprint is a vault note
• Daily-rhythm integration. This is a "run once, refine occasionally" skill, not part of the daily loop
• Auto-detection of publication/streaming gates from memory or identity. Phase 0 asks the user; auto-detection is deferred until a convention crystallizes

Implementation hints

• Pattern references in this repo: weekly-planning (Q&A + scout + scribe) and meeting-sync (Q&A with approval gates + archivist + scribe per spin-off). Both are good shape templates for the orchestration needed here.
• prism's existing remit (paradox / hidden patterns / unnamed concepts / sharp contrasts) covers Phase 3's synthesis pass without modification — confirm during implementation.
• Phase 1 ingest should accept pasted raw material (transcripts, highlights, notes) and feed it to archivist for pattern mining alongside the user's vault.
• Long session length is a real risk. Consider checkpointing the draft after Phase 2 or 3 so the user can resume across sessions.

Acceptance criteria

• plugins/athena-notes/skills/find-perspective/SKILL.md exists and is registered
• Triggering the skill on a vault with prior brand artifacts surfaces them in Phase 0 and offers a refine branch (not just from-scratch)
• Phase 2 produces a research summary per named thinker (worldview + key concepts), via sage
• Phase 3 lands on a mission statement the user explicitly approves before continuing
• Phase 5 writes a single vault note in the documented blueprint structure, via scribe
• Synthetic blank-slate run produces a coherent blueprint without prior artifacts
• Real-data run on existing brand artifacts produces a refinement (not a duplicate) of those artifacts

Open questions

None outstanding at scoping time.