Dianel555/DSkills / skills/agent-wiki
agent-wiki
UnverifiedIncremental LLM-friendly wiki generator for Obsidian note vaults. Use when: (1) Building wiki from notes, (2) Ingesting notes to wiki, (3) Obsidian LLM wiki, (4) Incremental knowledge base management. Triggers: 'build wiki from notes', 'ingest notes to wiki', 'Obsidian LLM wiki', 'incremental knowledge base'.
scripts/agent_wiki/quality.py
View on GitHub"""Quality tier scoring for wiki topics.
Computes deterministic structural metrics and assigns a five-tier rating
(stub/basic/standard/rich/premium) based on content richness. No LLM calls,
no network access, no I/O side effects.
"""
from __future__ import annotations
import re
import unicodedata
from typing import Any
from .config import nfc as _nfc
_LIST_RE = re.compile(r"^(?:[-*+]|\d+[.)\]])\s")
def _is_markdown_image_only(stripped: str) -> bool:
"""```` with nothing but punctuation after the closing paren."""
if not (stripped.startswith("![") and ")" in stripped):
return False
_, _, tail = stripped.partition(")")
return not any(c.isalnum() for c in tail)
def _is_non_prose(stripped: str) -> bool:
"""Heading, list, quote, table, comment, or embed-only line."""
return (
stripped.startswith(("#", ">", "|", "<!--"))
or bool(_LIST_RE.match(stripped))
or (stripped.startswith("![[") and stripped.endswith("]]"))
or _is_markdown_image_only(stripped)
)
def _count_cjk_and_latin(text: str) -> tuple[int, int]:
r"""Count CJK characters and Latin words in NFC-normalized text.
CJK characters: EAW in {W, F} AND category starts with L or N.
Excludes wide/fullwidth punctuation, symbols, emoji.
Halfwidth katakana (EAW=H) is normalized to fullwidth via NFKC first.
Latin words: after replacing wide/fullwidth chars with spaces,
count word runs via [^\W_]+ pattern.
Returns (cjk_chars, latin_words).
"""
# Apply NFC normalization
text = _nfc(text)
# Apply NFKC to normalize halfwidth katakana to fullwidth
text = unicodedata.normalize("NFKC", text)
cjk_count = 0
for ch in text:
eaw = unicodedata.east_asian_width(ch)
cat = unicodedata.category(ch)
# CJK: wide/fullwidth AND letter/number category
if eaw in ('W', 'F') and cat[0] in ('L', 'N'):
cjk_count += 1
# Replace wide/fullwidth chars with spaces to prevent merge
latin_text = ""
for ch in text:
eaw = unicodedata.east_asian_width(ch)
if eaw in ('W', 'F'):
latin_text += " "
else:
latin_text += ch
# Count word runs (letters/digits, excluding underscores)
words = re.findall(r'[^\W_]+', latin_text, re.UNICODE)
latin_count = len(words)
return cjk_count, latin_count
def compute_metrics(body: str) -> dict[str, Any]:
"""Compute structural quality metrics from topic body.
Metrics:
- sections: count of ## to ###### ATX headings (excluding level-1)
- evidence_lines: count of lines starting with "> "
- prose_chars: NFC character length of paragraph lines (retained for transparency)
- prose_weight: script-aware prose measure (10 × cjk_chars + 16 × latin_words)
- cjk_chars: count of CJK ideographs (EAW W/F + category L/N)
- latin_words: count of Latin/other word runs
- has_image: bool, body contains image embed
- has_lead: bool, first non-blank line is a paragraph
Returns dict with all metrics.
"""
lines = body.splitlines()
sections = 0
evidence_lines = 0
prose_chars = 0
cjk_total = 0
latin_total = 0
has_image = False
has_lead = False
in_fence = False
first_content_line_checked = False
seen_h1 = False
for line in lines:
stripped = line.strip()
# Toggle fence state
if stripped.startswith("```") or stripped.startswith("~~~"):
in_fence = not in_fence
continue
# Skip fenced content
if in_fence:
continue
# Skip blank lines
if not stripped:
continue
# Skip the first level-1 heading (topic title)
if not seen_h1 and re.match(r"^#\s+\S", stripped):
seen_h1 = True
continue
# Check for has_lead (first non-blank, non-fenced line after optional h1)
if not first_content_line_checked:
first_content_line_checked = True
# It's a paragraph if it's NOT a heading, list, quote, table, comment, or image
is_heading = stripped.startswith("#")
is_list = re.match(r"^[-*+]\s", stripped) or re.match(r"^\d+[.)\]]\s", stripped)
is_quote_or_table = stripped.startswith(">") or stripped.startswith("|")
is_comment = stripped.startswith("<!--")
is_wikilink_image = stripped.startswith("![[") and stripped.endswith("]]")
is_markdown_image = stripped.startswith("![") and ")" in stripped
if not (is_heading or is_list or is_quote_or_table or is_comment or
is_wikilink_image or is_markdown_image):
has_lead = True
# Count sections (## to ######)
if re.match(r"^#{2,6}\s+\S", stripped):
sections += 1
continue
# Count evidence lines
if stripped.startswith("> "):
evidence_lines += 1
continue
# Check for images
if not has_image:
# Obsidian embed: ![[filename.ext]]
obsidian_embed_match = re.search(r"!\[\[.+?\.(png|jpg|jpeg|gif|webp|svg|bmp)\]\]", stripped, re.IGNORECASE)
if obsidian_embed_match:
has_image = True
# Markdown image: 
markdown_image_match = re.match(r"!\[.*?\]\(.+?\)", stripped)
if markdown_image_match:
has_image = True
# Count prose characters and compute prose_weight
if not _is_non_prose(stripped):
prose_chars += len(_nfc(stripped))
cjk, latin = _count_cjk_and_latin(stripped)
cjk_total += cjk
latin_total += latin
prose_weight = 10 * cjk_total + 16 * latin_total
return {
"sections": sections,
"evidence_lines": evidence_lines,
"prose_chars": prose_chars,
"prose_weight": prose_weight,
"cjk_chars": cjk_total,
"latin_words": latin_total,
"has_image": has_image,
"has_lead": has_lead,
}
def compute_tier(body: str, source_count: int = 0, *, metrics: dict[str, Any] | None = None) -> str:
"""Assign quality tier based on metrics and source grounding.
Uses effective_prose = prose_weight + 500*source_count for tier gates.
Tiers (top-down first-match):
- premium: sections >= 6 AND effective_prose >= 3000 AND evidence_lines >= 3
- rich: sections >= 4 AND effective_prose >= 1500 AND (evidence_lines >= 1 OR has_image)
- standard: sections >= 2 AND effective_prose >= 600
- basic: effective_prose >= 200 OR sections >= 1
- stub: otherwise
Args:
body: Topic body markdown
source_count: Number of deduplicated sources (default 0)
metrics: Precomputed ``compute_metrics(body)`` to avoid a second pass
Returns tier string.
"""
if metrics is None:
metrics = compute_metrics(body)
sections = metrics["sections"]
prose_weight = metrics["prose_weight"]
evidence_lines = metrics["evidence_lines"]
has_image = metrics["has_image"]
# Compute effective prose with source grounding bonus
effective_prose = prose_weight + 500 * source_count
# Premium
if sections >= 6 and effective_prose >= 3000 and evidence_lines >= 3:
return "premium"
# Rich
if sections >= 4 and effective_prose >= 1500 and (evidence_lines >= 1 or has_image):
return "rich"
# Standard
if sections >= 2 and effective_prose >= 600:
return "standard"
# Basic (requires some prose OR structure; sources alone insufficient)
if (effective_prose >= 200 and prose_weight > 0) or sections >= 1:
return "basic"
# Stub
return "stub"
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |