Agents

agent
Security Audit
Fail
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Fail
  • exec() — Shell command execution in node_modules/tokenx/dist/cli.mjs
  • rm -rf — Recursive force deletion command in skills/get-pr-comments/scripts/get-pr-comments.sh
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

My $HOME/.agents folder

README.md

Agents

This repository stores my $HOME/.agents folder for AI agent use.

The agent files live at the root of this repository, which is cloned to $HOME/.agents.

Clone into $HOME/.agents

Bash:

# Optional: back up any existing ~/.agents directory
mv "$HOME/.agents" "$HOME/.agents.backup.$(date +%Y%m%d%H%M%S)"

# Clone this repository directly into ~/.agents
git clone -b main https://github.com/ByronScottJones/Agents.git "$HOME/.agents"

PowerShell (Windows):

# Optional: back up any existing ~/.agents directory
if (Test-Path "$HOME\.agents") {
  $timestamp = Get-Date -Format "yyyyMMddHHmmss"
  Rename-Item -Path "$HOME\.agents" -NewName ".agents.backup.$timestamp"
}

# Clone this repository directly into ~/.agents
git clone -b main https://github.com/ByronScottJones/Agents.git "$HOME\.agents"

If you already cloned the repo elsewhere, you can link it instead:

Bash:

ln -s /path/to/Agents "$HOME/.agents"

PowerShell (Windows):

# Create a symlink (requires admin or developer mode on Windows 10+)
New-Item -ItemType SymbolicLink -Path "$HOME\.agents" -Target "C:\path\to\Agents" -Force

Link into AI coding tools

Many tools work best when they can see the repo contents through a stable path such as $HOME/.agents or a symlink to it.

Recommended pattern

flowchart TD
    A[Shared Agents repo\n$HOME/.agents]:::shared

    subgraph Claude[Claude Code]
        C1[$HOME/.claude]
        C2[$HOME/.claude/agents]
    end

    subgraph Codex[Codex]
        X1[$HOME/.codex]
        X2[$HOME/.codex/agents]
    end

    subgraph Gemini[Gemini CLI]
        G1[$HOME/.gemini]
        G2[$HOME/.gemini/agents]
    end

    subgraph ChatGPT[ChatGPT]
        H1[$HOME/.chatgpt]
        H2[$HOME/.chatgpt/agents]
    end

    subgraph Copilot[Copilot]
        P1[$HOME/.copilot]
        P2[$HOME/.copilot/agents]
    end

    subgraph Other[Other tools]
        O1[$HOME/.antigravity_ide]
        O2[$HOME/.antigravity_ide/agents]
    end

    C1 --> C2
    X1 --> X2
    G1 --> G2
    H1 --> H2
    P1 --> P2
    O1 --> O2

    A -. symlink target .-> C2
    A -. symlink target .-> X2
    A -. symlink target .-> G2
    A -. symlink target .-> H2
    A -. symlink target .-> P2
    A -. symlink target .-> O2

    classDef shared fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px;

Example layout: each tool keeps its own home directory, and the agents subdirectory inside that tool points back to the shared $HOME/.agents repo.

Use a tool-specific parent directory for that tool's own files, then link the shared agent guidance into a common subdirectory such as agents. This lets each tool keep its own settings, cache, and state while sharing the same guidance files.

Bash:

#!/usr/bin/env bash

# Resolve shared Agents source directory:
# 1) Prefer ~/.agents if present
# 2) Otherwise prompt until a valid directory is entered
resolve_agents_src() {
  if [ -d "$HOME/.agents" ]; then
    printf '%s\n' "$HOME/.agents"
    return 0
  fi

  local input
  while :; do
    read -r -p "Enter the full path to your Agents repo: " input
    if [ -d "$input" ]; then
      printf '%s\n' "$input"
      return 0
    fi
    echo "Path does not exist or is not a directory: $input"
  done
}

# Create link only if missing; never replace existing paths
ensure_agents_link() {
  local parent_dir="$1"
  local link_path="$parent_dir/agents"
  local target="$2"

  mkdir -p "$parent_dir"
  if [ ! -e "$link_path" ]; then
    ln -s "$target" "$link_path"
    echo "Created: $link_path -> $target"
  else
    echo "Skipped (already exists): $link_path"
  fi
}

AGENTS_SRC="$(resolve_agents_src)"

ensure_agents_link "$HOME/.antigravity_ide"  "$AGENTS_SRC"
ensure_agents_link "$HOME/.chatgpt"  "$AGENTS_SRC"
ensure_agents_link "$HOME/.claude"  "$AGENTS_SRC"
ensure_agents_link "$HOME/.codex"   "$AGENTS_SRC"
ensure_agents_link "$HOME/.copilot" "$AGENTS_SRC"
ensure_agents_link "$HOME/.gemini"  "$AGENTS_SRC"

PowerShell (Windows):

# Resolve shared Agents source directory:
# 1) Prefer $HOME\.agents if present
# 2) Otherwise prompt until a valid directory is entered
function Resolve-AgentsSource {
  $defaultPath = Join-Path $HOME ".agents"
  if (Test-Path $defaultPath -PathType Container) {
    return $defaultPath
  }

  do {
    $inputPath = Read-Host "Enter the full path to your Agents repo"
    if (-not (Test-Path $inputPath -PathType Container)) {
      Write-Host "Path does not exist or is not a directory: $inputPath"
    }
  } until (Test-Path $inputPath -PathType Container)

  return $inputPath
}

# Create link only if missing; never replace existing paths
function Ensure-AgentsLink {
  param(
    [Parameter(Mandatory = $true)][string]$ParentDir,
    [Parameter(Mandatory = $true)][string]$Target
  )

  if (-not (Test-Path $ParentDir -PathType Container)) {
    New-Item -ItemType Directory -Path $ParentDir | Out-Null
  }

  $linkPath = Join-Path $ParentDir "agents"
  if (-not (Test-Path $linkPath)) {
    New-Item -ItemType SymbolicLink -Path $linkPath -Target $Target | Out-Null
    Write-Host "Created: $linkPath -> $Target"
  }
  else {
    Write-Host "Skipped (already exists): $linkPath"
  }
}

$agentsSrc = Resolve-AgentsSource

$AgentDirNames = @(
    ".antigravity_ide",
    ".chatgpt",
    ".claude",
    ".codex",
    ".copilot",
    ".gemini"
)

foreach ($dirName in $AgentDirNames) {
    Ensure-AgentsLink -ParentDir (Join-Path $HOME $dirName) -Target $agentsSrc
}

Grant default access so tools stop asking every time

The exact setting name varies by tool, but the goal is the same: set the workspace, instructions path, or trusted project root once so the tool can reuse it on future runs.

Claude Code

  • Add the repository root or .claude/agents to Claude Code's trusted workspace/project list if it supports one.
  • If Claude Code supports a default instructions or policy path, point it at $HOME/.claude/agents/AGENTS.md or the repo root.
  • Keep the symlink in place so Claude can access the shared guidance without prompting for a new location each time.
  • If Claude Code remembers approved directories, approve the parent directory that contains the agents link so the full path stays available.

Codex

  • Configure Codex to open the repository root or the tool-specific parent directory that contains agents.
  • If Codex supports a default workspace or trusted directory, set it once to the parent directory such as $HOME/.codex.
  • If Codex supports a custom instructions file, point it at $HOME/.codex/agents/AGENTS.md.
  • Approve or trust the parent directory so future sessions can reuse the same location without a new prompt.

Gemini CLI

  • Start Gemini CLI from the repository root or from a directory that already has access to the agents symlink.
  • If Gemini CLI supports a saved workspace or project directory, set it to $HOME/.gemini or the repo root once.
  • If Gemini CLI supports default instruction files, point it at $HOME/.gemini/agents/AGENTS.md.
  • Trust the parent directory so Gemini can reuse the shared guidance without asking again.

Other agent tools

  • Look for a setting named workspace, project root, trusted directory, instructions path, or prompt file.
  • Set it to the repo root or the tool-specific parent directory containing the agents symlink.
  • If the tool supports persistent approvals, approve the parent directory rather than just a single file.
  • If the tool supports startup arguments, pass the shared guidance path once and save that configuration for reuse.

Guidance Layout

  • AGENTS.md is always active and routes work to topic files.
  • Load only the topic files that match the task; load multiple files when a task
    spans languages or frameworks.
  • CUA_TUI_Specification.md applies only when CUA compliance is explicitly
    requested.

Validation

Bash:

npx --yes markdownlint-cli2 '**/*.md' '#.git/**'

PowerShell (Windows):

npx --yes markdownlint-cli2 '**/*.md' '#.git/**'

Skills

Skills are on-demand workflows under skills/<name>/SKILL.md. Load a skill
when its trigger matches the task; topic files remain the source of policy and
the skill provides the execution workflow.

Documentation and operations

Skill Purpose
generate-adr Draft architecture decision records.
generate-service-documentation Create service records and operational documentation.
incident-triage Triage incidents using bounded operational evidence.
prepare-release Prepare release notes and promotion readiness.
review-configuration Audit configuration safety and precedence.
validate-documentation Check Markdown, links, routing, and stale references.

Engineering and security

Skill Purpose
address-pr-comments Implement and validate pull request feedback.
build-and-scan-container-image Build, scan, and generate SBOMs for images.
execute-database-migration Plan, test, and execute schema migrations.
install-dev-tools Audit and install the coding tools the guidance requires (macOS, Windows, Ubuntu).
get-pr-comments Retrieve pull request review feedback.
remediate-dependency-vulnerabilities Scan and remediate vulnerable dependencies.
validate-and-deploy-iac Validate, plan, and safely apply infrastructure as code.
validate-and-deploy-kubernetes Validate and safely deploy Kubernetes manifests.
validate-github-actions Audit GitHub Actions syntax, permissions, and pinning.
verify-service-health Verify deployment readiness and service health.

TWG workflows

Skill Purpose
twg Use TWG whenever Atlassian/company context would help.
twg-agentic-search Deep internal research with Rovo.
twg-artifacts Share or update a local file as an Atlassian Artifact.
twg-code-review Additional code-review context.
twg-confluence Confluence content, spaces, and authoring.
twg-context-discovery Dependency maps, repos, and OOO catch-ups.
twg-engineering-work Code/repo discovery, PRs, and contributors.
twg-jira Jira workitems, issues, projects, boards, sprints, and writes.
twg-jira-resolve-merged-work Stale Jira work with merged PRs.
twg-operational-health Incidents/on-call, handoffs, Assets, and risk.
twg-responsibility-routing Owners, SMEs, approvers, and escalation.
twg-space-creation Create or clone Confluence spaces.
twg-status-rollups Project, goal, launch-readiness, and leadership rollups.

Reviews (0)

No results found