ai-knowledge-harvester

Stars1

Forks0

UpdatedApril 12, 2026 at 17:53

Harvests AI documentation into a centralized private knowledge repo using a project-first taxonomy. Two modes: (1) Repo scan - collects AI docs (specs, prompt logs, CLAUDE.md files, gap analyses, kickoff prompts, agent configs) from a source repo. (2) Ad-hoc ingest - captures pasted content exported from ChatGPT, Claude web, Clo.ai, or any AI interface. Triggers on phrases like 'harvest AI docs from', 'move my AI docs to', 'collect my prompt logs', 'sync AI docs to knowledge repo', 'organize my ai-docs folder', 'save this conversation', 'store this export', 'add this to my knowledge repo', or any request to consolidate AI documentation across repositories and AI tools into a single searchable private store. Also trigger when the user pastes a large markdown block that looks like an AI conversation export or asks to file away content from an AI session.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Source

jackzhaojin

jackzhaojin/ai-builder-kit

View GitHub Repository View Creator Repositories

Download

Related occupationsSOC

Based on SOC occupation classification

Software DevelopersComputer and Mathematical Occupations·SOC 15-1252

File Explorer

5 files

SKILL.md

readonly

More from this repository

same repository

playwright-demo-video

jackzhaojin/ai-builder-kit

Generate polished demo videos (MP4) from Playwright specs with AI voiceover, captions, and music. Use when the user says "demo video", "playwright demo", "generate demo", "caption extraction", "voiceover pipeline", "freeze frame merge", or "auto-discover demo" — or wants to: create demo videos for web projects, extract captions from Playwright specs, generate TTS voiceover with ElevenLabs, merge video with freeze-frame timing, add background music, or auto-discover project features to generate demo specs.

2026-06-111

chat-conversation-logger

jackzhaojin/ai-builder-kit

Log the current Claude.ai chat conversation as a timestamped, downloadable markdown file ready for an AI second brain. Captures the user's prompts verbatim — every word, plus attachment types — summarizes Claude's responses concisely, and notes every tool, artifact, and MCP call. Use this skill whenever the user types '/log', 'log this conversation', 'log this session', 'save conversation log', 'convo log', 'archive this chat', or any similar phrasing — even casually. This skill is manually invoked only: do NOT trigger on your own initiative without an explicit user request, even if the conversation seems like it might be useful to archive.

2026-05-241

conversation-logger

jackzhaojin/ai-builder-kit

Log AI coding agent conversation sessions with timestamps in structured markdown. Supports Claude Code, OpenAI Codex CLI, and GitHub Copilot. Use when the user says 'log this session', 'convo log', or wants to create a record of prompts and responses for a project. Auto-detects which agent is running and reads timestamps from the correct history file.

2026-05-241

jack-git-commit

jackzhaojin/ai-builder-kit

Structured git commit skill for all agents and humans. Use when: committing code changes, the user says "commit", "/jack-git-commit", a worker completes a task and needs to commit, or any agent needs to record work in git. Generates Conventional Commit messages with traceable metadata footers (Goal, Step, Worker) so git log serves as a self-documenting work ledger. Enforces atomic commits, staged-only policy, and NEVER pushes.

2026-05-151

drawio

jackzhaojin/ai-builder-kit

Generate draw.io diagrams (`.drawio` XML files) from detailed user prompts, in Jack's brand style. Use when the user asks to create, visualize, draw, or diagram architectures, workflows, pipelines, systems, or processes in draw.io / diagrams.net / XML form. Requires a description of what to diagram — does not analyze codebases automatically. Ask clarifying questions if diagram type, complexity, or subsystems are unclear. Every diagram is exported to PNG via the draw.io CLI and visually validated in a render-view-fix loop before delivery.

2026-04-271

competing-pr-triage

jackzhaojin/ai-builder-kit

Compare two pull requests or two branches that were produced by different agents attempting the same goal, and recommend which one to adopt. Use when the user asks to "compare PR X and PR Y", "which branch is better", "two agents gave me different PRs for the same task", "triage competing PRs", or any situation where multiple worktrees/branches solved (or attempted) the same goal in parallel and a decision is needed on which to merge or continue from. Produces a functional-requirements table with per-requirement verdicts, a file-level scope audit (extras check), a branch recommendation, and a concrete follow-up plan for what the agent should do after switching to the chosen branch.

2026-04-111

name	ai-knowledge-harvester
description	Harvests AI documentation into a centralized private knowledge repo using a project-first taxonomy. Two modes: (1) Repo scan - collects AI docs (specs, prompt logs, CLAUDE.md files, gap analyses, kickoff prompts, agent configs) from a source repo. (2) Ad-hoc ingest - captures pasted content exported from ChatGPT, Claude web, Clo.ai, or any AI interface. Triggers on phrases like 'harvest AI docs from', 'move my AI docs to', 'collect my prompt logs', 'sync AI docs to knowledge repo', 'organize my ai-docs folder', 'save this conversation', 'store this export', 'add this to my knowledge repo', or any request to consolidate AI documentation across repositories and AI tools into a single searchable private store. Also trigger when the user pastes a large markdown block that looks like an AI conversation export or asks to file away content from an AI session.
disable-model-invocation	true

AI Knowledge Harvester

Collect AI documentation into a centralized private ai-knowledge repository. Two modes, same output: classified, frontmattered markdown in a two-tier project taxonomy.

Capabilities

Capability	Description
Repo scan	Scan a source repo for AI docs (CLAUDE.md, specs, prompt logs, gap analyses, kickoff prompts) and copy them into the knowledge repo
Ad-hoc ingest	Capture pasted content from ChatGPT, Claude web, Clo.ai, Copilot, or any AI tool and file it into the knowledge repo
Monorepo support	Two-tier taxonomy: `{github-repo}/{sub-project}` — handles monorepos and simple repos alike
Auto-classify	Classify documents as `spec`, `prompt-log`, `working-doc`, or `harness-prompt` based on content signals
Frontmatter	Add standardized YAML frontmatter to every document for future indexing
README manifests	Create/update README.md at each folder level (1 level deep only) with Last Modified dates, sorted newest-first
Safe staging	Stage changes for review — never commit or push automatically
Non-destructive	Source files are kept in place by default — harvest is a copy, not a move

Parameters

Parameter	Required	Default	Description
`source_repo`	Mode 1 only	—	Path to the repo with AI docs
`knowledge_repo`	No	`{source_repo}/../ai-knowledge`	Path to the private ai-knowledge repo
`project_name`	No	GitHub remote repo name	Top-level project (the GitHub repo name)
`sub_project`	No	Same as `project_name` for simple repos, or detected sub-project name for monorepos	Second tier in taxonomy

For Mode 2 (ad-hoc), also determine doc_type, title, source_tool, and why_private from content or by asking the user. See references/ad-hoc-ingest.md.

Document Types

Type	Folder	Priority	Signals
`spec`	`specs/`	Primary	"specification", "requirements", "architecture", "decisions" — stable reference docs
`prompt-log`	`prompt-logs/`	Primary	Session transcripts, conversation exports, "→ Response / → Action" patterns
`working-doc`	`working-docs/`	Primary	Gap analyses, kickoff prompts, CLAUDE.md, AGENTS.md, operational guides. Default when uncertain.
`harness-prompt`	`harness-prompts/`	Secondary	AI-generated prompts, agent configuration templates, harness prompt files in `src/prompts/`. Optional — harvest but deprioritize.

Two-Tier Taxonomy

The primary key is always the GitHub repo name. Within it, a second tier for sub-projects.

Simple repo: {repo}/{repo}/ — repeat the name (e.g., ai-builder-kit/ai-knowledge-harvester/)
Monorepo: {repo}/{sub-project}/ — one folder per sub-project (e.g., jack-dev-server-configs/generic-harness-v1/)

ai-knowledge/
├── README.md
├── projects/
│   ├── README.md
│   ├── {github-repo}/                    # Top tier: GitHub repo name
│   │   ├── README.md                     # Repo overview, why private, sub-project list
│   │   └── {sub-project}/               # Second tier: sub-project (or repo name repeated)
│   │       ├── README.md                 # Sub-project overview, type folder list
│   │       ├── specs/
│   │       │   ├── README.md
│   │       │   └── YYYY-MM-DD-{seq}-{slug}.md
│   │       ├── prompt-logs/
│   │       │   ├── README.md
│   │       │   └── YYYY-MM-DD-{seq}-{slug}.md
│   │       ├── working-docs/
│   │       │   ├── README.md
│   │       │   └── YYYY-MM-DD-{seq}-{slug}.md
│   │       └── harness-prompts/          # Optional — only if harness prompts exist
│   │           ├── README.md
│   │           └── YYYY-MM-DD-{seq}-{slug}.md
│   └── {another-repo}/
│       └── ...
└── shared/
    ├── README.md
    └── skills/

Each README describes its folder's purpose and lists immediate children only — never recurses deeper. For full README templates and frontmatter spec, see references/taxonomy.md.

Core Workflow

Mode 1: Repo Scan

Validate source_repo and knowledge_repo paths exist
Determine project name from GitHub remote, detect sub-projects (monorepo or simple)
Discover AI docs — scan ai-docs/, local-ai-docs/, local-only/, local/, nested subdirectories
Present classified list to user, wait for confirmation
Create project folder structure, write files with frontmatter
Create/update README.md files at each level (1 level deep)
Stage and commit changes — never push
Source files stay in place — only offer cleanup if the user explicitly asks

For detailed steps, scan patterns, and exclusion rules: see references/repo-scan.md.

Mode 2: Ad-Hoc Ingest

Read pasted content — detect source tool, classify type, identify project + sub-project
Ask the user if project, sub-project, type, or title is ambiguous — don't guess
Clean up formatting, prepend frontmatter with source_tool field
Write to correct type folder, update READMEs, stage and commit — never push

For detailed steps, source detection hints, and example interactions: see references/ad-hoc-ingest.md.

Edge Cases

Situation	Action
File already exists in destination	Compare dates — newer overwrites (with note), older skips. If dates are equal, keep the larger (more comprehensive) file. Log every collision decision
Ambiguous file type	Default to `working-doc`, tell the user
Monorepo with sub-projects	Use two-tier: `{repo}/{sub-project}/` — detect sub-projects from directory structure
Simple repo (not a monorepo)	Repeat repo name: `{repo}/{repo}/`
Multiple docs on same date	Use sequence numbers in filename (`01`, `02`) to preserve chronological order
No git history for date	Use today's date, add `date_note` in frontmatter
Large markdown files (up to 50MB)	Harvest them — storage is free. Only skip files >50MB
Binary files / images / PDFs	Skip, mention them, ask if user wants to handle separately
Pasted content with no context	Ask project and type — don't assume
Mixed-topic paste	Ask: file as one doc or split?
Non-markdown paste	Convert to clean markdown, preserve substance
"Save this conversation"	Summarize current session as prompt-log or working-doc, ask project
`local-ai-docs/` found in repo	Harvest everything in it — this folder exists specifically for AI docs awaiting harvest. Recommend adding to `.gitignore` if not already ignored

Safety Rules

Never push. Stage and commit, but never push to remote.
Never modify source files during harvest. Read-only access to source repos during harvest.
Never delete source files unless explicitly asked. Harvest is a copy, not a move. Source files stay in place by default.
Never silently overwrite. Report every file action, especially collision resolutions (log both versions' dates and sizes).
Ask when uncertain. Project, type, and title must be confirmed when ambiguous.

Why This Skill Exists

For the design rationale — why GitHub+markdown over Notion/CosmosDB, why project-first taxonomy, the open source philosophy, and what comes next — see references/spirit.md.