cd /news/ai-agents/show-hn-taskless-turn-agents-md-and-… · home › topics › ai-agents › article
[ARTICLE · art-174319] src=taskless.io ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

Show HN: Taskless, turn AGENTS.md and PR comments into runnable checks

Solo developer Jakob released Taskless, an MIT-licensed CLI that reads AGENTS.md files, PR review comments and fixup commits and converts recurring issues into deterministic ast-grep and Vale code checks committed to the repository, so no model runs at check time. The tool is available via `npx @taskless/cli` and an "onboard with Taskless" agent prompt, and it ships with anonymous telemetry that can be disabled by setting DO_NOT_TRACK=1 or TASKLESS_TELEMETRY_DISABLED=1. Taskless published sample reports from the Effect-TS/effect, getsentry/sentry and n8n-io/n8n repositories, with Jakob noting n8n has more AI-driven code than the other two and shows how quickly agent-copied problems propagate without determinism in place.

key takeaway

Taskless converts recurring AGENTS.md and PR review issues into committed ast-grep and Vale checks that run without a model, keeping enforcement deterministic after the agent's one-time onboarding pass.

read2 min views6 publishedNov 22, 2026

Hey HN, I'm Jakob, the solo dev and founder behind Taskless.

I got tired of my AGENTS.md files being ignored or forgotten a half dozen turns into conversations. I also remembered reading someone here on HN saying we weren't using nearly enough syntax trees in our code checks. This is the result.

Taskless gives you powerful syntax tree tools with human verifiable fixtures. The CLI is MIT licensed, so running the CLI, onboarding, rule writing, and running the underlying ast-grep & Vale checks are all local. You can give it a try with:

npx @taskless/cli

and then asking your agent to "onboard with Taskless".

Under the hood, the CLI works with your agent to read your PR review comments, fixup commits, and agent files. It takes things that keep recurring and turns them into the simplest code-level enforcement that can prevent the issue going forward. By choosing code over an LLM for the runtime, there's no model once code is generated. The checks can then be committed to your codebase and enforced. I'm particularly proud of how determinism is mixed into the onboarding flow. Any time the CLI can give the agent information without letting the agent wandering the codebase, the CLI does. It results in a more consistent experience with far fewer errors. Anonymous telemetry can be disabled outright by setting DO_NOT_TRACK=1 or TASKLESS_TELEMETRY_DISABLED=1 in your environment (it collects CLI version, OS, the repo's org/owner, and a bit about the repository it's running in).

When the agent's done, it creates a report you can share around to show everyone there's a better way to control the output of agentic coding tools. I thought I'd include a few here from some of my favorite repos.

I'd love to hear what you're doing today to keep agents behaving, or if you do kick the tires, I'd love the honest feedback.

Comments URL: [https://news.ycombinator.com/item?id=50603421](https://news.ycombinator.com/item?id=50603421)

Points: 2

── more in #ai-agents 4 stories · sorted by recency
── more on @taskless 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/show-hn-taskless-tur…] indexed:0 read:2min 2026-11-22 · —