Submit Paragent to awesome-lists once the repo is presentable

[myselfsiddharth]

What

Submit Paragent to two curated lists. Gated — do not submit until the preconditions below are met.

Awesome-lists are a slow trickle of qualified traffic rather than a spike, but the entry is permanent and compounds. The cost is one small PR each.

Do not submit yet — preconditions

Both lists reject low-effort entries, and a rejected submission is awkward to retry with the same maintainer.

• docs: make root README a visitor landing page, preserve internal one #133 merged — a maintainer clicking through today lands on "Pre-seed. Thesis unproven. Gate number pending" with no idea what the project does
• Publish to npm so Paragent is installable without cloning #134 landed (npm) — right now there is no install path at all; package.json is private: true. An entry whose only instruction is "clone the repo" reads as unfinished
• Record and embed demo GIF #137 landed (demo GIF) — nice-to-have, not blocking, but it is what makes a list entry convert once someone clicks

The honest test: would you merge this entry if a stranger submitted it? Submit when the answer is yes.

Target 1 — angrykoala/awesome-browser-automation (~632 stars)

Best fit. Smaller list, tightly on-topic, actively maintained.

Their contribution guidelines allow self-submission outright: "Additions of your own tools or resources are welcome, as long as they are awesome, documented, and functional." No star minimum.

Section: AI — their guidelines route anything "primarily focused on AI, or to be used with AI (e.g. Agents, MCP, automation with prompts)" there.

Format: [tool](link) - Description. — alphabetical, capital letter, full stop, no trailing slash on the link.

Draft entry (goes under P):

[Paragent](https://github.com/DevToolie/Paragent) - Records a browser agent's successful run and replays it as a deterministic script with per-step assertions, calling the model back only to repair steps that break.

Requirements to satisfy:

• One PR, this suggestion only
• Correct alphabetical position in the AI section
• PR description explains why it is awesome, not just what it is
• Link is direct, no trailing slash
• If submitted by an automated agent, their guidelines require a joke in the PR message. Not optional — it is in their rules, and it is how they spot agents that did not read them

Target 2 — e2b-dev/awesome-ai-agents (~29k stars)

Much bigger reach, weaker category fit. The list catalogs autonomous agents themselves; Paragent is infrastructure that agents plug into. Expect low marginal traffic given the number of existing entries, and a real chance of being passed over as out-of-scope.

Submit by PR or via their form. No CONTRIBUTING.md; conventions are inferred from the README.

Format: a per-project block with ### Category, ### Description (bulleted), ### Links. Alphabetical, correct category.

• Pick the least-wrong category — likely "Build your own" rather than a general-purpose agent category
• Description bullets carry no metrics (see below)

Constraint that applies to both

No performance claims. Per docs/INTEGRITY-AUDIT.md A-05 and B-02, replay cost is [PENDING TRACK-1]. Describe the mechanism — "replays without the model in the loop" — never a saving. An external list is exactly where an unmeasured number would get quoted back at us, and we would not be able to source it.

Also worth considering later

• Jenqyang/Awesome-AI-Agents (~1.2k) and kaushikb11/awesome-llm-agents (~1.6k) — both pushed within the last day, so actively maintained. Lower priority than the two above.

Not in scope

Show HN and other launch-moment channels. Those are a separate, larger effort with different timing — awesome-lists are the slow-burn complement, not a substitute.

Open questions / what I could not verify

• Whether either maintainer considers a pre-seed project with an unproven thesis in scope at all. Neither list says anything about project maturity beyond the low-effort rule.
• Whether e2b-dev/awesome-ai-agents prefers the form over PRs in practice — the README offers both without saying which gets processed.
• Whether the "Related tools" section of awesome-browser-automation is a better home than AI, if the maintainer reads Paragent as infrastructure rather than an AI tool.

[myselfsiddharth]

Blocked — a precondition that reads as met is not.

This issue gates on:

• Publish to npm so Paragent is installable without cloning #134 landed (npm) — right now there is no install path at all; package.json is private: true. An entry whose only instruction is "clone the repo" reads as unfinished

#134 is closed as completed and #155 shipped the paragent binary, so the checkbox looks satisfiable. It is not — the package was made publishable and then never published:

$ npm view paragent version
npm error 404 Not Found - GET https://registry.npmjs.org/paragent - Not found

The README's Quick Start tells visitors to run npx paragent record --fixture --out trajectory.json, which fails for anyone who has not cloned the repo. Filed as #167.

Submitting now would be worse than not submitting: a maintainer evaluating the entry follows the README's first command and hits a 404. Both target lists reject low-effort entries, and this issue already notes that a rejected submission is awkward to retry with the same maintainer.

Hold until #167 resolves — either by publishing, or by replacing the npx Quick Start with a path that works. The other two preconditions (#133 landing page, #137 demo GIF) are genuinely met.