Reuse
Borrowing it
Nothing to install: this file belongs to dotdevdotdev/agentwire-dev. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dotdevdotdev/agentwire-dev/main/.claude/skills/agentwire-config/SKILL.mdgit clone --depth 1 https://github.com/dotdevdotdev/agentwire-devWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/dotdevdotdev/agentwire-dev/agentwire-config)<a href="https://agentmods.dev/skills/dotdevdotdev/agentwire-dev/agentwire-config"><img src="https://agentmods.dev/badge/skills/dotdevdotdev/agentwire-dev/agentwire-config/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/dotdevdotdev/agentwire-dev/agentwire-config"><img src="https://agentmods.dev/badge/skills/dotdevdotdev/agentwire-dev/agentwire-config.svg" alt="Reviewed on agentmods" width="80" height="20"></a> Per session 89 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,496 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits How audits are shown
- NVIDIA SkillSpector warn
SkillSpector: 4 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Privilege Escalation · line 57 Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.Fix: Remove references to credential paths. Use environment variables or secrets managers. For docs, use placeholder paths (e.g., /path/to/config). Never load .env or token files in production code paths.
- high Privilege Escalation · line 60 Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.Fix: Remove references to credential paths. Use environment variables or secrets managers. For docs, use placeholder paths (e.g., /path/to/config). Never load .env or token files in production code paths.
- high Privilege Escalation · line 164 Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.Fix: Remove references to credential paths. Use environment variables or secrets managers. For docs, use placeholder paths (e.g., /path/to/config). Never load .env or token files in production code paths.
- medium Data Exfiltration · line 93 Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
Origin original No closer match found in the catalogue.
Token cost
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00089 | $0.03496 |
| Opus 5.5 | $0.00036 | $0.01398 |
| Sonnet 5.5 | $0.00018 | $0.00699 |
| Haiku 4.5 | $0.00009 | $0.00350 |
Security
Grade A, and why
agentwire-config scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
command: "curl -sf ..." # for command (exit 0 = healthy) .claude/skills/agentwire-config/SKILL.md · 254 lines
How it starts
The opening of the file, as written. The whole thing — 254 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AgentWire Config (~/.agentwire/config.yaml)
Layout of ~/.agentwire/
| File | Purpose |
|---|---|
config.yaml |
Main config (see structure below) |
machines.json |
Remote machines registry |
scripts/ |
Machine-specific helper scripts (TTS management, startup, etc.) |
voices/ |
Custom TTS voice samples |
uploads/ |
Uploaded images for cross-machine sharing |
artifacts/ |
Agent-generated HTML for artifact windows |
wiki/ |
LLM-maintained knowledge base (Karpathy LLM Wiki pattern) |
logs/ |
Audit logs for damage-control |
Per-session config (posture, roles, voice) lives in .agentwire.yml in each project directory (see agentwire-project-config skill).
Machine Scripts (~/.agentwire/scripts/)
Each machine has a ~/.agentwire/scripts/ directory for machine-specific helper scripts (TTS management, startup hooks, service wrappers, etc.). This is the standard location — agents should look here first and put new scripts here.
Scripts in ~/bin/ should symlink to ~/.agentwire/scripts/ so they're callable from PATH but the source of truth is in one place.
These scripts are not managed by agentwire — they're local to each machine and not version controlled. They exist because different machines have different roles (GPU server runs TTS, Mac runs the portal, etc.) and need different glue scripts.
config.yaml Structure
server:
host: "127.0.0.1" # default; "0.0.0.0" allows LAN/phone access and requires the auth token (see SECURITY.md)
port: 8765
activity_threshold_seconds: 3 # Seconds before session considered idle
ssl:
cert: "~/.agentwire/cert.pem"
key: "~/.agentwire/key.pem"
# Auth token: unset = use ~/.agentwire/portal.token (auto-generated on first
# non-loopback start; print/rotate with `agentwire portal token [--rotate]`).
# Set a string to override the file; "" disables auth (loopback binds only —
# the portal refuses to start on 0.0.0.0 with auth disabled).
# auth_token: ""
# Extra browser origins allowed on state-changing requests (exact
# scheme://host[:port]). The portal's own origin and localhost always pass.
# Needed when fronting with Cloudflare Tunnel:
allowed_origins: [] # e.g. ["https://portal.example.com"]
projects:
dir: "~/projects"
worktrees:
enabled: true
suffix: "-worktrees"
auto_create_branch: true
copy_files: [".env", ".agentwire.yml", ".agentwire.tasks.yml"] # gitignored files seeded into each new worktree
# (git worktree add only checks out tracked files,
# so .env/secrets/local config don't carry over —
# add ".env.local", ".envrc", etc. as needed).
# Keep .agentwire.yml/.agentwire.tasks.yml gitignored: a TRACKED
# copy means worktree runs use the committed version (HEAD) and
# silently ignore live edits — see agentwire-project-config skill.
tts:
backend: "default" # tier: default (in-process Kokoro, zero setup — ~200MB model
# auto-downloads on first portal start; speechSynthesis covers
# the wait) | custom (self-hosted shim at url)
url: "http://localhost:8100" # custom tier only — shim endpoint
default_voice: "dotdev"
voices_dir: "~/.agentwire/voices" # Custom voice samples for cloning
instructions: "" # free-text prompt passed through to the shim
options: # opaque JSON passed to the shim; the bundled shim reads:
backend: kokoro # engine: kokoro | chatterbox | chatterbox-streaming | zonos-transformer | zonos-hybrid
exaggeration: 0.5 # Voice expressiveness (0-1, Chatterbox)
cfg_weight: 0.5 # CFG weight (0-1, Chatterbox)
timeout: 60
stt:
backend: "default" # TIER (where transcription happens): default (portal-owned
# in-process Moonshine — bundled, auto-downloads on first boot,
# no setup; falls back to browser SpeechRecognition while it
# warms up or on py3.14+) | cloud (portal → hosted OpenAI-
# compatible transcription API, no shim daemon) | custom
# (self-hosted shim at url)
engine: "auto" # ENGINE (which model the self-hosted shim loads): auto | moonshine |
# whisper. Orthogonal to backend — used only by `agentwire stt start/serve`.
# `{backend: custom, engine: whisper}` = boot shim AND run faster-whisper.
moonshine_model: "moonshine/base" # moonshine engine only — ONNX model id (moonshine/tiny | moonshine/base)
model: "base" # whisper engine only — faster-whisper/openai-whisper model (tiny → large-v3)
url: "http://localhost:8101" # custom tier only — shim endpoint (also the `agentwire stt` port)
cloud: # cloud tier only — all fields optional, defaults shown
base_url: "https://api.openai.com/v1" # any OpenAI-compatible endpoint (Groq, Mistral, speaches, ...)
model: "gpt-4o-mini-transcribe"
api_key_env: "OPENAI_API_KEY" # NAME of the env var holding the key — the key itself
# never lives in config and never reaches the browser;
# portal refuses to start if the var is unset
language: "" # optional ISO-639-1 hint
timeout: 30
silence_prepend_ms: 0 # prepend silence if your backend clips the first syllable
instructions: "" # free-text hint passed through to the shim
options: {} # opaque JSON passed to the shim (language hints, vocab biasing, ...)
corrections: {} # post-transcription find/replace, e.g. {"agent wire": "agentwire"}
agent:
command: "claude --dangerously-skip-permissions"
dev:
source_dir: "~/projects/agentwire-dev" # agentwire source for TTS/STT venv
services: # Where services run (for multi-machine setups)
portal:
machine: null # null = local
port: 8765
session_name: "agentwire-portal" # tmux session name
tts:
machine: "gpu-server" # or null for local
port: 8100
session_name: "agentwire-tts"
stt:
session_name: "agentwire-stt"
custom: # User-defined service sessions — autostart on portal launch AND
# `agentwire up`, health-checked by the portal watchdog, shown in
# the portal's Services column. Manage with `agentwire services ...`.
# The notifications bridge is a built-in registry entry (override
# by defining a service with its name).
- name: "agent-brain" # tmux session name (required)
project: "~/projects/brain" # project dir; defaults to dev source dir
autostart: true # boot on portal launch / `agentwire up` (default true)
roles: "brain" # optional; overrides project .agentwire.yml
posture: "bypass" # optional; posture override
restart: on-failure # never | on-failure | always (watchdog respawn
# with 30s..10m exponential backoff; default on-failure;
# `agentwire services down` always sticks)
healthcheck: # optional; defaults to tmux_session/60s
kind: tmux_session # tmux_session | http | command
url: "http://..." # for http (2xx = healthy)
command: "curl -sf ..." # for command (exit 0 = healthy)
interval: 60 # seconds between watchdog checks
- "simple-service" # string shorthand = name only, all defaults
executables: # Override executable paths (optional, auto-detected by default)
ffmpeg: "/opt/homebrew/bin/ffmpeg"
whisperkit-cli: "/opt/homebrew/bin/whisperkit-cli"
hs: "/opt/homebrew/bin/hs"
agentwire: "~/.local/bin/agentwire"
uploads:
dir: "~/.agentwire/uploads"
max_size_mb: 10
cleanup_days: 7
artifacts:
dir: "~/.agentwire/artifacts"
max_size_mb: 10
wiki:
dir: "~/.agentwire/wiki" # Wiki vault location
portal:
url: "https://localhost:8765"
channels: # Outbound-only notifications. Only email + quo ship.
# Keys are env-only: RESEND_API_KEY / QUO_API_KEY in ~/.agentwire/.env
# (docs/wiki/security/secrets.md) — never in config.yaml.
email:
from_address: "Echo <[email protected]>"
default_to: "[email protected]"
banner_image_url: "https://yourdomain.com/images/banner.png"
echo_image_url: "https://yourdomain.com/images/echo.png"
echo_small_url: "https://yourdomain.com/images/echo-small.png"
logo_image_url: "https://yourdomain.com/images/logo.png"
quo:
from_number: "+1234567890" # E.164 or phone number ID (PNxxx)
default_to: "+0987654321"
scheduler:
autostart: true # Start the scheduler daemon when the portal boots (default: true)
dispatch_cooldown: 60 # Seconds between task dispatches (default: 60)
dispatch_max_runtime: 14400 # Watchdog ceiling per dispatch in seconds; a hung ensure is killed and the task marked timeout (default: 4h, 0 disables)
usage_limit: # Usage-limit recovery watchdog (docs/wiki/usage-limit-recovery.md)
enabled: true # Master switch for dialog detection/parking (default: true)
exclude_sessions: [] # Session names never auto-parked (gates NEW parks only)
session_context: # Context-bloat observability (Phase 0, observe-only — issue #442)
warn_remaining_pct: 20 # Flag a session when its REMAINING context drops to/below this %.
# The Claude Code bar shows headroom, not usage, so LOW = bloated.
# Surfaced via `agentwire list --context` and MCP `sessions_context`.
worktree: # `agentwire worktree <name>` orchestration (WorktreeConfig).
# Distinct from projects.worktrees above (the legacy
# project/branch layout).
# PRECEDENCE (#705): a project's .agentwire.yml `worktree:`
# block (dir/base) overrides these for that repo; a
# per-invocation --base flag beats both. Chain: flag →
# project .agentwire.yml → this global block → built-ins.
worktree_dir: ~/worktrees # Root for worktrees, nested per project:
# <worktree_dir>/<project>/<name>/ (mirrors ~/projects/)
default_base: develop # Base branch new worktrees fork from. OMIT to derive from
# the repo's actual default branch (origin/HEAD, fallback to
# current branch) — no hardcoded 'main'. --base always wins.
default_project: ~/projects/my-repo # Repo used when --project is omitted AND cwd isn't in a
# git repo. Otherwise --project / the git root of cwd is used.
naming: "{user}/{slug}" # Optional branch-name template for NEW branches. Placeholders:
# {name} (verbatim), {slug} (slugified), {user} (OS login).
# Omit → branch == name verbatim. Only the git branch is
# templated; the tmux session name stays {project}-{name}.
session:
# No global default-role: a session's ROLE is derived from its spawn verb
# (new → orchestrator or worker depending on branch, worktree → worker by
# default, spawn → worker), then any --roles / .agentwire.yml roles: stack
# on (worker) or replace (orchestrator) it. TOPOLOGY (worktree vs pane/main)
# separately picks WHICH worker etiquette file (worker-worktree vs worker).
# See resolve_roles.
inject_soul: true # Append the bundled 'soul' personality role to every human-facing
# session (appended last for recency weight). Headless roles
# (worker, task-runner, notifications) and soul/soul-* sessions
# are excluded automatically; per-session opt-out: --no-soul on new/dev
beta:
# Opt-in gates for features that SHIP on main but stay off until asked for.
# Every flag defaults to false, and "off" means ABSENT — not merely dormant:
# a gated feature's role-prompt lines AND its MCP tool-description prose are
# stripped before a model sees them, so a user who never enabled it pays no
# tokens for it. Only a real YAML boolean turns one on — `<flag>: "false"`
# (quoted) stays OFF, deliberately. `agentwire doctor` reports each flag's
# state either way. Config-only (no CLI verb); env override works via
# AGENTWIRE_BETA__<FLAG>=true.
# (no flags currently defined)
Changes
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago Changed · -2 lines 35177dd3cc7b
- 2mo ago First seen · 256 lines · 89 tokens per session scan A 7a0d3dfbeabc
agentwire-config is a skill published in the GitHub repository dotdevdotdev/agentwire-dev (23 stars, last pushed yesterday), licensed Apache-2.0. It adds 89 tokens to every session and 3,496 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Related
agentwire-config compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| agentwire-configdotdevdotdev/agentwire-dev | Skill | 23 | yesterday | Apache-2.0 | A | 89 | Claude Code |
| tmux-via-mcpbnomei/tmux-mcp | Skill | 57 | 3mo ago | MIT | A | 39 | Claude Code, Codex |
| tmux-buffer-explorerbnomei/tmux-mcp | Skill | 57 | 3mo ago | MIT | A | 33 | Claude Code, Codex |
| datamolsynthetic-sciences/openscience | Skill | 4.3k | 2d ago | Apache-2.0 | A | 67 | Claude Code, Codex |
| clinicaltrials-databasesynthetic-sciences/openscience | Skill | 4.3k | 2d ago | Apache-2.0 | A | 47 | Claude Code, Codex |
| glycoengineeringsynthetic-sciences/openscience | Skill | 4.3k | 2d ago | Apache-2.0 | A | 75 | Claude Code, Codex |
| esmsynthetic-sciences/openscience | Skill | 4.3k | 2d ago | Apache-2.0 | A | 86 | Claude Code, Codex |