User repository on GitHub
Dynamic, confidence-aware Agent Skill routing with TypeSafe Jev and Tink
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Tink Route
CLI that routes a task to one specialist skill from a Tink library using TypeSafe Jev, verifies the winner with tink mount, then prints it inline or as a mounted path.
- Stars
- 5
- Language
- Python
- License
- MIT
- Published by
- Community
- Author
- jon-devlapaz
- Last verified
- Oct 2026
About this project
Routes a task to one specialist skill from your Tink library, verifies it with tink mount, and prints it, using TypeSafe Jev for the decision. Then store the routing key (paste it at the silent prompt): Requires Python 3.11+ and the tink CLI on PATH. The key file is the reliable place for the key: agent harnesses often start tools from non-interactive shells that never read your shell profile, so an exported variable can silently be missing there.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- Decide-only --pick prints the winner and confidence without mounting, writing, or emitting a receipt.
- Skills up to --inline-max characters print inline; larger skills mount under .tink/.active/ without truncation.
- Whole-library search by default; --skillset restricts strictly and can suggest a non-delivered hint from another shelf.
- Machine-readable --json covers routed, delivered, no-skill, and error states with confidence and reason slugs.
Watch out
Requires Python 3.11+, the tink CLI on PATH, and a TYPESAFE_API_KEY for routing decisions. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Grep finds text matches but cannot judge fit, verify approval, or check digests. This asks Jev to gate, rank, and rerank candidates, then only prints content that passes tink mount verification.
It fails open: it prints one plain sentence naming the fix, prints no skill content after a refusal, and exits non-zero. Exit 1 means no skill applies, exit 2 means delivery or usage failed, with a stable reason slug in JSON.
It never reads AGENTS.md, so stage shelf rules neither scope nor break routing. Receipts append one JSON line per delivery when you pass --receipt PATH, using append with a file lock, and symlinked receipt paths are refused.
Use --pick for decide-only output with confidence and runner-up margin, --skillset to restrict to one shelf with a non-delivered hint, and --model, --threshold, rerank, and fits flags listed under Advanced in --help.
pipx install git+https://github.com/jon-devlapaz/tink-route.git, since it is not on PyPI; it needs Python 3.11+ and the tink CLI on PATH, and you store the key at ~/.config/tink-route/typesafe_api_key with mode 600 rather than an exported variable a harness may miss.
Three steps: a gate question on whether the task is a specialised workflow, a rank where Jev picks the most load-bearing candidate or abstains, and a rerank of the top three with SKILL.md excerpts and per-skill fit scores.
tink-route picks and verifies one skill for a task you name and prints it, while tink-inject routes needs and injects approved, verified skills into an agent's context at the right moment so the agent never learns skills exist.
More Jev resources like this
25GitHub stars
Agent-oriented playbook for using Jev as a cheap, fast decision layer that triages, classifies, scores, and gates work before expensive browser, research, retry, or subagent steps.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
Bodila51/muse-jev-playbookPython
Bodila51/muse-jev-playbookPython14GitHub stars
Opt-in Hermes plugin that routes each turn to one skill: a wide Jev Choice plus request checks gate a narrow fits round, injecting a single ignorable hint line.
Jev Agents & MCP#community#jevbest#catalog-import
6GitHub stars
LangChain deepagents middleware that replaces SkillsMiddleware with per-turn judging, loading only the SKILL.md files a turn needs so large catalogs stay out of the prompt.
Jev Agents & MCP#community#jevbest#catalog-import
3GitHub stars
A Codex Skill that hands Jev the task summary plus the host's real model list, so Jev picks a model and reasoning effort for an independent subtask and Codex executes with it.
Jev Agents & MCP#community#awesomejev#catalog-import
2GitHub stars
A Python CLI that asks Jev which of four coding agents should take a task, optionally waits for your approval, then opens that app or its website and pastes the prompt in.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
AABBAASS1/jev-routerPython
AABBAASS1/jev-routerPython3GitHub stars
Agent Skill that sends large tool output and CI logs to Jev and hands your expensive model only the few lines that matter, shipped with a stdlib-only caller, a doctor command, and a plan mode that keeps open-ended work on your LLM.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
lazniak/jevskillPython
lazniak/jevskillPythonFrom the Jev community
Posts from builders shipping with Jev right now.