persona-profile-from-textlisted

# Persona Profile from Text Build a **mirror profile** of one person from their own words: what they keep returning to, what they say they value, where the words they choose quietly disagree with the life they describe. Every line of the output quotes the actual source. This is a mirror held up to a person, not a verdict passed on them. The method comes from Shiva's design note (the bundled copy is `references/00-method.md`; the canonical note is in the wiki at `_raw/2026-06-22-psychological-profiling-from-text-design.md`). It has two named use modes, set in STEP 0. **Golden rule: questions, not verdicts. Frameworks are a lens, never a score.** You may reason with validated frameworks (Big Five, Schwartz values, affect) to organize your thinking, but you never print a trait name, a number, a percentile, or a type code. A surprising read is offered as an invitation to reflect, never as a diagnosis. A false or preachy read is the fastest way to lose the person's trust. **Paths:** when this skill loads, its base directory is announced. Capture it as `$SKILL_DIR` and use it for bundled files. Reads of `references/*.md` may use bare relative paths, but Bash runs must use the absolute base (the runtime CWD is the user's project, not this skill): `SKILL_DIR="<announced base directory>"`, then `bash "$SKILL_DIR/scripts/lint.sh" <file...>`. --- ## STEP 0: Which mode (decide first) Produce a profile in exactly ONE mode per request. - **Mirror mode.** The text belongs to a Rehea