persona-evaluator

Name: persona-evaluator
Author: acnlabs

40stars5forks

How to install persona-evaluator?

acnlabs/openpersona/persona-evaluator

$npx -y skills add acnlabs/OpenPersona --skill persona-evaluator

Installs into the current project.

Summary

Audit any OpenPersona (or peer LLM-agent) persona in three complementary modes: structural (CLI, deterministic, CI-friendly: 4 Layers × 5 Systemic Concepts × Constitution gate with role-aware severity), semantic white-box (LLM reads pack-content JSON and scores Soul-narrative qua

Files · 1

View on GitHub

SKILL.md

1	# persona-evaluator — Persona Quality Auditor
2
3	Score any OpenPersona persona pack against the 4+5 framework standard:
4	4 Layers (Soul · Body · Faculty · Skill) × 5 Systemic Concepts (Evolution · Economy · Vitality · Social · Rhythm) + Constitution compliance gate.
5
6	`persona-evaluator` reads `persona.json`, generated artifacts, and soul files to produce a structured 9-dimension report — calibrated to the OpenPersona quality standard, with role-aware severity and three modes for self / peer / black-box review.
7
8	---
9
10	## Quick Start
11
12	```bash
13	# Evaluate an installed persona (static / structural)
14	npx openpersona evaluate <slug>
15
16	# JSON output (for scripting or CI)
17	npx openpersona evaluate <slug> --json
18
19	# Save report to file (always JSON; --json not needed alongside --output)
20	npx openpersona evaluate <slug> --output report.json
21
22	# Embed evaluable persona content (Soul/character/behavior-guide) so an
23	# LLM evaluator (this skill, acting through an agent) can also judge
24	# quality semantically — not just structurally
25	npx openpersona evaluate <slug> --pack-content
26	```
27
28	## Choosing a mode
29
30	`persona-evaluator` runs in three complementary modes. Pick the mode based on what the user asks before reading the rest of this file.
31
32
33	\| User asks \| Mode \| How \| Confidence \|
34	\| ---------------------------------------------------- \| ----------------------------- \| -------------------------------------------------------------------- \| --------------------------- \|
35	\| "CI / gate persona quality" \| structural \| `npx openpersona evaluate <slug>` \| deterministic \|
36	\| "Polish review of my own pack" \| semantic white-box (self) \| `... evaluate <slug> --pack-content`, then apply rubric in self-mode \| high \|
37	\| "Peer-review a pack I have on disk" \| semantic white-box (peer) \| same command, peer-mode rubric \| high \|
38	\| "Review agent X" where X is remote / non-OpenPersona \| semantic black-box \| A2A handshake → consent + probe → passive, in that order \| mid (cap 8/10) or low (cap 6/10) \|
39
40
41	Structural is the default. Switch to semantic only when the user explicitly asks for narrative quality review (e.g. "evaluate me semantically", "self-review my pack", "qualitative audit"). Switch to black-box only when you cannot read the subject's `persona.json` on disk.
42
43	Sections below cover each mode in depth: structural ([What Gets Scored](#what-gets-scored)), semantic white-box ([Semantic Evaluation](#semantic-evaluation-llm-driven)), and semantic black-box ([Black-box Semantic Evaluation](#black-box-semantic-evaluation)).
44
45	## What Gets Scored
46
47	The structural CLI scores 9 dimensions + the Constitution gate. Severity (`strict` / `normal` / `lenient`) is set per dimension by the persona's declared role.
48
49
50	\| Layer / Concept \| Dimension \| Looks at \|
51	\| ---------------- \| ------------------------------ \| ------------------------------------------------------ \|
52	\| Soul \| identity, character, aesthetic \| `persona.json` Soul block + `soul/*.md` \|
53	\| Body \| environment, runtime \| hardware/runtime declaration \|
54	\| Faculty \| tools, capabilities \| declared

People who liked this also viewed

6 picks

Type

persona-evaluator

40stars5forks

View on GitHub

How to install persona-evaluator?

acnlabs/openpersona/persona-evaluator

$npx -y skills add acnlabs/OpenPersona --skill persona-evaluator

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# persona-evaluator — Persona Quality Auditor
2
3	Score any OpenPersona persona pack against the 4+5 framework standard:
4	4 Layers (Soul · Body · Faculty · Skill) × 5 Systemic Concepts (Evolution · Economy · Vitality · Social · Rhythm) + Constitution compliance gate.
5
6	`persona-evaluator` reads `persona.json`, generated artifacts, and soul files to produce a structured 9-dimension report — calibrated to the OpenPersona quality standard, with role-aware severity and three modes for self / peer / black-box review.
7
8	---
9
10	## Quick Start
11
12	```bash
13	# Evaluate an installed persona (static / structural)
14	npx openpersona evaluate <slug>
15
16	# JSON output (for scripting or CI)
17	npx openpersona evaluate <slug> --json
18
19	# Save report to file (always JSON; --json not needed alongside --output)
20	npx openpersona evaluate <slug> --output report.json
21
22	# Embed evaluable persona content (Soul/character/behavior-guide) so an
23	# LLM evaluator (this skill, acting through an agent) can also judge
24	# quality semantically — not just structurally
25	npx openpersona evaluate <slug> --pack-content
26	```
27
28	## Choosing a mode
29
30	`persona-evaluator` runs in three complementary modes. Pick the mode based on what the user asks before reading the rest of this file.
31
32
33	\| User asks \| Mode \| How \| Confidence \|
34	\| ---------------------------------------------------- \| ----------------------------- \| -------------------------------------------------------------------- \| --------------------------- \|
35	\| "CI / gate persona quality" \| structural \| `npx openpersona evaluate <slug>` \| deterministic \|
36	\| "Polish review of my own pack" \| semantic white-box (self) \| `... evaluate <slug> --pack-content`, then apply rubric in self-mode \| high \|
37	\| "Peer-review a pack I have on disk" \| semantic white-box (peer) \| same command, peer-mode rubric \| high \|
38	\| "Review agent X" where X is remote / non-OpenPersona \| semantic black-box \| A2A handshake → consent + probe → passive, in that order \| mid (cap 8/10) or low (cap 6/10) \|
39
40
41	Structural is the default. Switch to semantic only when the user explicitly asks for narrative quality review (e.g. "evaluate me semantically", "self-review my pack", "qualitative audit"). Switch to black-box only when you cannot read the subject's `persona.json` on disk.
42
43	Sections below cover each mode in depth: structural ([What Gets Scored](#what-gets-scored)), semantic white-box ([Semantic Evaluation](#semantic-evaluation-llm-driven)), and semantic black-box ([Black-box Semantic Evaluation](#black-box-semantic-evaluation)).
44
45	## What Gets Scored
46
47	The structural CLI scores 9 dimensions + the Constitution gate. Severity (`strict` / `normal` / `lenient`) is set per dimension by the persona's declared role.
48
49
50	\| Layer / Concept \| Dimension \| Looks at \|
51	\| ---------------- \| ------------------------------ \| ------------------------------------------------------ \|
52	\| Soul \| identity, character, aesthetic \| `persona.json` Soul block + `soul/*.md` \|
53	\| Body \| environment, runtime \| hardware/runtime declaration \|
54	\| Faculty \| tools, capabilities \| declared

People who liked this also viewed

6 picks

Type