Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/LilMGenius/polysona/llms.txt

Use this file to discover all available pages before exploring further.

The $interview / /interview command launches a deep psychology-based extraction session driven by the profiler agent. It is the first and most fundamental step in the Polysona pipeline — without a completed interview, no other skill has persona data to work with. Run it once to establish a baseline, or return to it at any time to continue deepening the record.

​
Invocation

$interview

​
What it does

The profiler agent follows a resume-first protocol on every invocation:
  • Reads personas/_active.md to determine the active persona directory (falls back to default).
  • Attempts to load personas/{active}/persona.md and locate the ## interview-log section.
  • If a prior log exists, it reads the latest entries and resumes from the last meaningful point — picking up the last framework covered, open GAP: hypotheses, and any unresolved tensions — rather than restarting from zero.
  • Confirms continuity in its first prompt, naming the prior framework before continuing the depth-spiral.
When no persona file exists, the profiler starts a fresh interview:
  • Opens with the McAdams narrative warmup: asks for life chapters, one high point, one low point, and one turning point.
  • Captures narrative language verbatim before moving on.
  • Transitions to Laddering only after narrative anchors are established.
Across both protocols, the following execution rules always apply:
  • All 10 psychology frameworks are exercised over the full interview flow (time allocation is not necessarily equal).
  • Contradictions are flagged inline using the GAP: format and written to the interview-log.
  • The session stays in extraction mode — no persona structuring happens here.
  • The log is append-only: historical entries are never overwritten.

​
Prerequisites

None for a first run. For a resumed session, personas/{active}/persona.md must already exist with an ## interview-log section.
If you have multiple personas, make sure personas/_active.md points to the correct persona directory before invoking $interview.

​
Output

All interview data is appended to: 
personas/{active}/persona.md   ← interview-log section
Each run adds timestamped log entries. Contradictions appear inline as: 
GAP: [description of the contradiction or unresolved tension]
No structuring or synthesis is performed at this stage — that happens in downstream steps.

​
Guardrails & Notes

The profiler operates in extraction-only mode. Do not expect structured persona artifacts (nuance.md, accounts.md) to be generated here — those are produced by the Polysona structuring engine after the interview.
  • The append-only rule is strict: deleting or editing past log entries will break resume detection.
  • GAP: rows are log entries marking contradictions — they are not framework names and should not be treated as such.
  • If the active persona path resolves to a directory that does not exist, the profiler falls back to default.
  • Framework coverage is tracked per run; the profiler will prioritise frameworks not yet represented in the existing log.
The $interview skill is executed by the profiler agent. See the Profiler agent page for details on its 10-framework interview flow, defense-bypass prompts, and log extraction behaviour.

Build docs developers (and LLMs) love

Get started for freeTalk to us