skills/wiki-builder/SKILL.md
Start, structure, and grow a persistent research wiki indexed in pro-workflow's SQLite knowledge base. Each wiki is a folder of markdown pages with provenance, plus a shadow FTS5 index so any session can recall it. Use when the user says "start a wiki", "add to wiki", "compile a page", "wiki on X", or wants a long-lived knowledge base on a topic, paper, product, person, project, or codebase.
npx skillsauth add rohitg00/pro-workflow wiki-builderInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Persistent knowledge base for any topic. Markdown on disk + SQLite FTS5 shadow index.
~/.pro-workflow/wikis/<slug>/ — default, never committed<project>/.claude/wikis/<slug>/ — pass --scope project, committableBoth register in the same ~/.pro-workflow/data.db.
| Flavor | Use for |
|--------|---------|
| research | ongoing topic exploration |
| paper | one-paper deep dive |
| domain | broad subject area |
| product | product/tool KB |
| person | researcher/founder dossier |
| organization | company/lab profile |
| project | internal project KB |
| codebase | symbol/file-aware KB tied to a repo |
| incident | post-mortem KB |
<slug>/
├── wiki.config.md # purpose, audience, page types, style, auto_research block
├── raw/ # untouched source material (PDFs, scrapes, transcripts)
├── wiki/
│ └── index.md # entry point, hand-curated TOC
├── derived/ # generated artifacts (surveys, charts, summaries)
├── prompts/ # per-task prompts (compile-page, lint, query)
├── logs/maintenance-log.md
└── sources.md # one row per source: id | url | title | hash | fetched_at
Flavor adds folders: wiki/papers, wiki/concepts, wiki/people, wiki/products, wiki/timelines, wiki/questions.
node $SKILL_ROOT/scripts/wiki-cli.js init <slug> --title "X" --flavor research [--scope project] [--root <path>]
node $SKILL_ROOT/scripts/wiki-cli.js list
node $SKILL_ROOT/scripts/wiki-cli.js page <slug> <rel-path> --title "X" [--type concept|paper|person|...] [--from-file path]
node $SKILL_ROOT/scripts/wiki-cli.js reindex <slug>
node $SKILL_ROOT/scripts/wiki-cli.js info <slug>
init runs init_wiki.sh (mirrors dair layout) AND registers the wiki in SQLite. page writes markdown + upserts FTS row.
wiki.config.md of the target wiki before any compile.wiki/ must cite a row in sources.md (one citation = one source row).wiki-cli.js page so FTS index stays in sync.logs/maintenance-log.md per change.wiki/index.md if new top-level page.tool-use-benchmarks.md, not 2026-05-08-notes.md).> SPECULATION: block.Wikis with private: true in config never get fetched from web sources by wiki-research-loop. Local raw/ only.
Phase 3.3.0 ships builder + query only. Loop arrives in 3.3.1. To prep, wiki.config.md may include:
auto_research:
enabled: false # flip in 3.3.1
max_pages_per_run: 5
max_depth: 3
budget_usd: 0.50
fetchers: [web, arxiv, github]
See templates/ for wiki.config.md, index.md, prompt files. init_wiki.sh copies these into the new wiki root.
devops
SkillOpt-flavored offline training loop for any SKILL.md. Treats accumulated learn-rule corrections as training trajectories, proposes bounded patches via an optimizer LLM, gates each candidate against a held-out validation set built from the user's own past corrections, and ships only candidates that demonstrably improve the score. Inspired by Microsoft SkillOpt's ReflACT pipeline (rollout → reflect → aggregate → select → update → evaluate) adapted to pro-workflow's SQLite store. Use when a skill has accumulated 8+ learn-rule rows and the user wants the skill itself to get better, not just longer.
tools
Prevent destructive operations using Claude Code hooks. Three modes — cautious (warn on dangerous commands), lockdown (restrict edits to one directory), and clear (remove restrictions). Uses PreToolUse matchers for Bash, Edit, and Write.
development
Complete AI coding workflow system. Orchestration patterns, 18 hook events, 5 agents, cross-agent support, reference guides, and searchable learnings. Works with Claude Code, Cursor, and 32+ agents.
data-ai
Analyze permission denial patterns and generate optimized alwaysAllow and alwaysDeny rules. Use when permission prompts are slowing you down or after sessions with many denials.