Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add onsager-ai/dev-skills --skill worktree-devproxygit clone --depth 1 https://github.com/onsager-ai/dev-skillsWrote 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/onsager-ai/dev-skills/worktree-devproxy)<a href="https://agentmods.dev/skills/onsager-ai/dev-skills/worktree-devproxy"><img src="https://agentmods.dev/badge/skills/onsager-ai/dev-skills/worktree-devproxy/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/onsager-ai/dev-skills/worktree-devproxy"><img src="https://agentmods.dev/badge/skills/onsager-ai/dev-skills/worktree-devproxy.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00274 | $0.02713 |
| Opus 5.5 | $0.00110 | $0.01085 |
| Sonnet 5.5 | $0.00055 | $0.00543 |
| Haiku 4.5 | $0.00027 | $0.00271 |
Grade C, and why
worktree-devproxy scanned grade C with 2 findings 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 1mo 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
| `wt rm` fails `Permission denied` | Root-owned files written by containers into the bind mount. The bundled `wt` already chowns via a throwaway alpine container and falls back to `rm -rf` + `git worktree prune`. | Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
3. Verify: `curl -s --noproxy '*' localhost:8080/api/version` returns the Traefik version; `curl localhost:8000` returns 404 (no routes yet — correct). How it starts
The opening of the file, as written. The whole thing — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.
worktree-devproxy
One shared Traefik + one compose stack per git worktree. Each instance is reachable at http://<worktree>.<repo>.localhost:8000; a laptop reaches it through a single SSH tunnel. Nothing else touches the host network.
laptop browser → SSH tunnel (:8000) → Traefik (127.0.0.1 only)
→ Host-header match → per-worktree compose stack (devproxy network)
Invariants (don't trade these away when adapting):
- No host ports except Traefik's own
127.0.0.1:8000(web) and127.0.0.1:8080(dashboard/API). Everything else flows over the shared external docker networkdevproxy. - Routing is Host-header only —
*.localhostresolves to loopback by OS convention, so there is no DNS, no TLS, and no port allocation table. - Worktree isolation is
COMPOSE_PROJECT_NAME(set by.envrcvia direnv): independent networks, volumes, and container names per worktree.
One-time machine setup
Prerequisites: docker (with compose plugin), git, direnv (hook it into the shell rc).
- Copy
references/dev-proxy.docker-compose.ymlto~/dev-proxy/docker-compose.yml,mkdir -p ~/dev-proxy/dynamic, anddocker compose up -d(run withCOMPOSE_PROJECT_NAME=dev-proxyor from a direnv-free shell — a stray project name from another repo's.envrccreates a second Traefik that loses the :8000 bind race). Two flags in it are load-bearing, learned the hard way:image: traefik:v3(not an old pin) — Traefik ≤ v3.3 hardcodes Docker API 1.24, which Docker Engine 29+ rejects outright.--providers.docker.network=devproxy— app containers sit on[default, devproxy]; without the pin Traefik dials the default network's IP it can't reach and every request 504s.
- Install
scripts/wtandscripts/devproxy-routeto~/bin/,chmod +x, make sure~/binis on PATH. - Verify:
curl -s --noproxy '*' localhost:8080/api/versionreturns the Traefik version;curl localhost:8000returns 404 (no routes yet — correct).
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 164 lines · 274 tokens per session scan C 9304bd7a041d
worktree-devproxy is a skill published in the GitHub repository onsager-ai/dev-skills (6 stars, last pushed 7d ago), licensed MIT. It adds 274 tokens to every session and 2,713 once invoked, about $0.0011 per session on Opus 5.5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
deployment-patterns
Deployment workflows, CI/CD pipeline patterns, Docker containerization, health checks, rollback strategies, and production readiness checklists for web applications. Use when setting up deployment infrastructure or planning releases.
deploy-akashic
A deployment procedure for an Akashic project that checks and merges an approved pull request, then submits the resulting code revision to update a Docker-based service. The host system handles stopping, moving, and restarting the service.
gear
Managing dependencies, CI/CD, advanced GitHub Actions workflows, containers, secrets, and operational config. Use for build, workflow, or environment work.
factory-clean
Use when the user asks to reclaim project disk space or clean up leftovers — inventories worktrees, containers, dependencies, and processes, reclaiming only user-approved items; read-only by default.
release
Used to release all toolbox, integrations, agents. Use when releasing a subproject to PyPI, Docker Hub, or when the user asks to release or publish.
archestra-dev-resource-cleanup
Reclaim unused Archestra development worktrees, local branches, processes, and OrbStack Kubernetes resources while preserving the active Tilt stack and unfinished work.