New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

First Run

by Abilityai in Abilityai/cornelius

First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this…

MITAuto-check: notesDatabases

Install First Run

skills CLI
$ npx skills add Abilityai/cornelius --skill first-run -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Abilityai/cornelius first-run --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Abilityai/cornelius.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/first-run .claude/skills/first-run && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
first-run
GitHub stars
109
Token cost
~2k tokens
SKILL.md length
746 words
Files
1
Skills in repo
54
Repo updated
First seen
Licence
MIT

At a glance

First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this…

  • Works in 6 steps: Is a bootstrap already running or done? → The Python environment → Rebuild the index once (before the daemon) → …
  • Tasks that involve Vector databases
  • SKILL.md covers Purpose, State Dependencies, Process and Refusals and limits, plus 1 more section
  • Calls python, python3 and pip; reaches download.pytorch.org

What it does

First Run is an agent skill from Abilityai/cornelius. First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this machine, runs one smoke search and reports what works. Idempotent - run it again any time to verify. On Trinity, the same bootstrap starts on its own at first boot; this playbook finishes it and proves it.

Its SKILL.md is about 2k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Databases, covering Vector databases, Embeddings and Local SEO. It works with Python. The repository describes itself as: AI-powered second brain template for Claude Code + Obsidian. The licence is MIT.

When your agent uses it

  • Tasks that involve Vector databases
  • Tasks that involve Embeddings
  • Tasks that involve Local SEO

Example prompts

  • “/first-run”

Requirements

  • Python 3
  • Pre-approved tools (allowed-tools): Bash, Read

Workflow steps

6 steps, taken from the step headings in SKILL.md.

  1. Is a bootstrap already running or done?
  2. The Python environment
  3. Rebuild the index once (before the daemon)
  4. The search daemon
  5. Verify with one smoke search
  6. Report

What it can do on your machine

Read from SKILL.md and the folder at commit b9bea90. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves these tools, so the agent can use them without asking each time:

    • Bash
    • Read

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python
    • python3
    • pip

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Hosts in commands or code, which the agent is likely to contact:

    • download.pytorch.org

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

First Run loads about 2k tokens when it runs. Until then it costs about 104 tokens; SKILL.md has 746 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~104
When it runs · the whole SKILL.md, loaded when a task matches
~2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
    allowed-tools: Bash, Read

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Abilityai/cornelius at commit b9bea90, republished under its MIT licence (© Abilityai). 746 words, ~2,017 tokens.

Download SKILL.mdSave it as .claude/skills/first-run/SKILL.md (or your agent's skills folder).
name
first-run
description
First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this machine, runs one smoke search and reports what works. Idempotent - run it again any time to verify. On Trinity, the same bootstrap starts on its own at first boot; this playbook finishes it and proves it.
allowed-tools
Bash, Read
user-invocable
true
argument-hint
[--verify-only]
metadata.version
1.2
metadata.created
2026-10-08
metadata.author
Ability.ai
metadata.changelog
1.2: Install in the foreground of one streaming Bash call, never detached - on Trinity v0.9.5 the agent-server orphan sweeper kills unowned background…

First Run

ℹ️ Print one line first: first-run v1.2 - recent: foreground install, never detached (orphan sweeper). Then proceed.

Purpose

A fresh copy of Cornelius has the whole seeded knowledge base and a prebuilt search index, but not the Python environment that reads it, and the index still carries the file paths of the machine that built it. Until both are fixed, /recall, /advise, /decide, /find-connections, /extract-insights and every other search-backed playbook return nothing or fail. This playbook takes the vault from "cloned" to "answers questions" in one command, and reports honestly what it found at every step.

Safe to run repeatedly: every step checks before it acts.

State Dependencies

SourceLocationReadWrite
Search engineresources/local-brain-search/ (venv, requirements.txt, run_daemon.sh, run_index.sh, run_search.sh)✓✓ (venv, daemon, index files)
Index manifestresources/local-brain-search/data/manifest.json✓via reindex
Bootstrap log and marker (Trinity)~/.trinity-bootstrap.log, ~/.trinity-bootstrap.done, ~/.trinity-bootstrap.lock✓✓
The vault$VAULT_BASE_PATH (default ./Brain)✓

Process

Work from the agent root (the folder that holds Brain/ and resources/). Keep every Bash call short (under 60 s): long installs run detached and are polled, so the platform's stall watchdog never fires.

Step 1 - Is a bootstrap already running or done?
bash
ls -la ~/.trinity-bootstrap.done ~/.trinity-bootstrap.lock 2>/dev/null; tail -5 ~/.trinity-bootstrap.log 2>/dev/null
  • .done present → the boot-time bootstrap finished; skip to Step 5 (verify).
  • .lock present and the log still growing → it is running; poll tail -3 ~/.trinity-bootstrap.log every 30 s until .done appears or the log stops for 2 minutes. A log that stops after "venv built" or "reindex start" with no process behind it means the boot-time bootstrap was swept (see Step 2) - continue with Step 2; every step checks before it acts, so nothing is redone.
  • Neither → continue with Step 2.
Step 2 - The Python environment
bash
cd resources/local-brain-search && ./venv/bin/python -c "import faiss, sentence_transformers, networkx; print('ok')" 2>&1 | tail -1

If that prints ok, skip to Step 3. Otherwise build it in the foreground of one Bash call, with output streaming (pip prints as it goes, so the platform's stall watchdog does not fire; on a 2-vCPU box with the CPU-only torch wheel this takes 1-2 minutes, ~1.5 GB):

bash
cd resources/local-brain-search && python3 -m venv venv && ./venv/bin/pip install --no-cache-dir --extra-index-url https://download.pytorch.org/whl/cpu -r requirements.txt 2>&1 | grep -v -E "Downloading|Using cached|^\s+━" ; ./venv/bin/python -c "import faiss, sentence_transformers; print('ok')"

Tell the user it is installing before you run it. On an error, paste the last 20 lines and stop - do not retry blindly. Do not run the install detached (nohup … &) on Trinity: the agent-server's orphan sweeper kills background processes it does not own ~90 s after container start, which is exactly how the boot-time bootstrap gets cut off mid-way - this playbook exists to finish the job inside an execution the platform owns.

Step 3 - Rebuild the index once (before the daemon)

The shipped index remembers the paths of the machine that built it. One rebuild reuses the shipped embeddings and rewrites the paths; it takes seconds to a couple of minutes and downloads the embedding model (~90 MB) on first use.

bash
cd resources/local-brain-search && ./run_index.sh 2>&1 | tail -8
Show full SKILL.md (302 more words)Show less
Step 4 - The search daemon
bash
cd resources/local-brain-search && timeout 120 ./run_daemon.sh start

"Daemon already running" is success. "started but not yet responding" → wait 20 s and run ./run_daemon.sh status. If the daemon will not come up, searches still work through the CLI fallback, only slower - do not block on it.

bash
cd resources/local-brain-search && BRAIN_READ_SCOPE=core,books,document-insights ./run_search.sh "decision under uncertainty" --limit 3 --json 2>/dev/null | python3 -c "import sys,json; d=json.load(sys.stdin); r=d.get('results',d); print(len(r),'results'); [print('-', x.get('title') or x.get('note') or x.get('path')) for x in r[:3]]"

(The scope is widened on purpose: the engine's default read scope is core, your own notes, which is nearly empty on a fresh copy; the seed lives in the reading pile. /recall and /advise mount the wider scope themselves.)

Three titles → the engine, the index and the seed are all there. Zero results → the index was not rebuilt (go back to Step 4) or the daemon is serving the old one (reload it). An import error → Step 2 did not finish.

Also confirm the orb data exists: ls -la resources/agent-visualization/data.json (on Trinity, .trinity/setup.sh copies it from the seed at boot; locally the Brain Orb does not apply).

Step 6 - Report

One short block, no prose padding:

First run - <date>
Python env     : ok (faiss, sentence-transformers, torch) | built now in <m> min | FAILED: <reason>
Search daemon  : running (pid <n>) | not responding
Index          : rebuilt now | already current - <notes> notes, <edges> edges (manifest.json)
Smoke search   : 3 results - <title 1> · <title 2> · <title 3>
Brain Orb data : present | n/a
Try next       : /recall <topic>   ·   /advise <a problem you are weighing>   ·   /decide <a choice with options>

If --verify-only was passed, run Steps 1, 2 (check only), 3 (status only), 5 and 6 and change nothing.

Refusals and limits

  • Never run the install in the foreground of a single Bash call; it would exceed the stall watchdog and look like a hang.
  • Never delete or rebuild the venv when the import check passes.
  • If the machine has under ~1.2 GB free (free -m), say so before installing: the install may be killed and the install should be retried after stopping other containers.
  • Does not touch the vault, the git state, or any credential.

Verification (for the author)

  • Fresh fork on Trinity v0.9.5, 4 GB droplet: /first-run completes in 4-8 minutes, /recall sunk cost then returns three cited notes.
  • Second run: every step reports "already", nothing is rebuilt, the report still shows three titles.

© Abilityai, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/first-run of Abilityai/cornelius.

Open the folder on GitHubat commit b9bea90

Compare with similar skills

First Run next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

First Run compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
First Run this skillAbilityai/cornelius109—~2kAutomated safety check: NotesMIT
FAISS Similarity SearchOrchestra-Research/AI-Research-SKILLs13k6 repos~1.3kAutomated safety check: PassMIT
DBoracle/skills876—~1.4kAutomated safety check: PassUPL-1.0
Chroma Vector DatabaseOrchestra-Research/AI-Research-SKILLs13k7 repos~2.3kAutomated safety check: PassMIT
Retail Product Search Agentgoogle/adk-recipes10k—~3kAutomated safety check: PassApache-2.0
Codebase Explorationgiancarloerra/SocratiCode3.3k1 repos~1.5kAutomated safety check: PassAGPL-3.0

Similar skills

  • FAISS Similarity Search

    Orchestra-Research/AI-Research-SKILLs

    Sets up FAISS for fast nearest-neighbor search over large collections of dense vectors, choosing between Flat, IVF, HNSW and product quantization indexes.

    13k GitHub starsUsed in 6 repos~1.3k tokens
    DatabasesAuto-check passed

  • DB

    oracle/skills

    Official

    Oracle Database guidance for SQL, PL/SQL, SQLcl, ORDS, Oracle Vector SDK, administration, app development, performance, security, migrations, and agent-safe database workflows.

    876 GitHub stars~1.4k tokensUpdated 2 days ago
    DatabasesAuto-check passed

  • Chroma Vector Database

    Orchestra-Research/AI-Research-SKILLs

    Shows how to store documents and embeddings in Chroma, query them by similarity with metadata filters, and persist them to disk for RAG and semantic search projects.

    13k GitHub starsUsed in 7 repos~2.3k tokens
    AI & LLM EngineeringAuto-check passed

  • Retail Product Search Agent

    google/adk-recipes

    Official

    Builds a retail product search agent on Google Cloud, from catalog ingestion into BigQuery and Vector Search to ADK scaffolding, evaluation and Cloud Run deployment.

    10k GitHub stars~3k tokensUpdated today
    AI & LLM EngineeringAuto-check passed

  • Codebase Exploration

    giancarloerra/SocratiCode

    Explore and understand codebases using SocratiCode semantic search, dependency graphs, and context artifacts.

    3.3k GitHub starsUsed in 1 repo~1.5k tokens
    DatabasesAuto-check passed

  • Pinecone Vector Database

    Orchestra-Research/AI-Research-SKILLs

    Shows how to use Pinecone, a managed vector database, for production RAG, semantic search and recommendations: indexes, upserts, queries, filters and namespaces.

    13k GitHub starsUsed in 5 repos~2k tokens
    DatabasesAuto-check passed

More from Abilityai/cornelius

All 54 skills in this repo

  • Nano Banana Image Generator

    Abilityai/cornelius

    Generate images using Google's Nano Banana (Gemini 2.5 Flash Image).

    109 GitHub stars~1.2k tokensUpdated yesterday
    Auto-check: notes

  • Benchmark Memory

    Abilityai/cornelius

    Systematic benchmarking framework for Local Brain Search memory system with LLM-as-judge scoring

    109 GitHub stars~3.1k tokensUpdated yesterday
    Auto-check: notes

  • Changelog Protocol

    Abilityai/cornelius

    Protocol for creating dated changelog files after significant agent sessions.

    109 GitHub stars~555 tokensUpdated yesterday
    Auto-check passed

  • Create Article

    Abilityai/cornelius

    Create long-form articles from knowledge base insights. An agent skill from Abilityai/cornelius.

    109 GitHub stars~3k tokensUpdated yesterday
    Auto-check: notes

  • Decide

    Abilityai/cornelius

    Structure a decision, not just advise on it. An agent skill from Abilityai/cornelius.

    109 GitHub stars~2.7k tokensUpdated yesterday
    Auto-check: notes

  • Epistemic Classification

    Abilityai/cornelius

    Framework for distinguishing research findings from hypotheses and speculative synthesis.

    109 GitHub stars~1.5k tokensUpdated yesterday
    Auto-check passed

Works with

Categories

Questions about First Run

What does First Run do?

First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this…. First Run is an agent skill from Abilityai/cornelius. First-install playbook for a fresh Cornelius - builds the local search engine (Python venv + FAISS + the embedding model), starts the search daemon, rebuilds the index so its stored paths match this machine, runs one smoke search and reports what works.

When should I use First Run?

First Run fits situations like: tasks that involve Vector databases; tasks that involve Embeddings; tasks that involve Local SEO.

How do I install First Run in Claude Code?

Run `npx skills add Abilityai/cornelius --skill first-run -a claude-code`. Or copy the skill folder (.claude/skills/first-run in Abilityai/cornelius) into .claude/skills/first-run in your project. Claude Code loads it when a task matches its description.

How do I install First Run in Codex?

Run `npx skills add Abilityai/cornelius --skill first-run -a codex`. Or copy the skill folder (.claude/skills/first-run in Abilityai/cornelius) into .agents/skills/first-run in your project. Codex loads it when a task matches its description.

Can I use First Run in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Abilityai/cornelius --skill first-run -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/first-run, .gemini/skills/first-run, .github/skills/first-run and .opencode/skills/first-run in your project.

What does First Run need to run?

Going by SKILL.md and its folder, First Run needs the command-line tools its instructions call (python, python3 and pip). Our summary lists: Python 3. Its frontmatter pre-approves these tools: Bash, Read.

Does First Run access the network?

SKILL.md names 1 domain. In commands or code: download.pytorch.org; the agent is likely to contact it when it follows the instructions. This is read from the text; nothing was executed.

Is First Run safe to install?

Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does First Run use?

First Run is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does First Run use?

About 2k tokens (SKILL.md is roughly 8.1k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to First Run?

Skills that share tags, products or a category with First Run: FAISS Similarity Search (Orchestra-Research/AI-Research-SKILLs, 13k stars), DB (oracle/skills, 876 stars), Chroma Vector Database (Orchestra-Research/AI-Research-SKILLs, 13k stars) and Retail Product Search Agent (google/adk-recipes, 10k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains First Run?

Abilityai (a GitHub organization) maintains it in Abilityai/cornelius, which has 109 GitHub stars. The repository holds 54 skills in this directory. The repository was last updated on October 8, 2026.

Source: Abilityai/cornelius on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.