@uge/payo 2.5.0 → 2.5.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +56 -31
- package/dist/index.js +188 -184
- package/package.json +12 -4
package/README.md
CHANGED
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
|
|
3
3
|
<img src="https://raw.githubusercontent.com/uttam-gelot/payo/main/assets/logo.png" alt="Payo" width="200" />
|
|
4
4
|
|
|
5
|
-
**
|
|
5
|
+
**One interview. Rules your AI can't skip.**
|
|
6
6
|
|
|
7
7
|
Payo interviews you about your stack, then writes the guidance files your AI
|
|
8
8
|
coding assistant reads — so Claude, Cursor, Copilot, and friends follow _your_
|
|
@@ -32,6 +32,7 @@ project's conventions instead of guessing.
|
|
|
32
32
|
- [Already have a project? Payo detects your stack](#already-have-a-project-payo-detects-your-stack)
|
|
33
33
|
- [What gets generated](#what-gets-generated)
|
|
34
34
|
- [What it asks about](#what-it-asks-about)
|
|
35
|
+
- [Code and PR reviews](#code-and-pr-reviews)
|
|
35
36
|
- [AI vs. template generation](#ai-vs-template-generation)
|
|
36
37
|
- [Bootstrap prompt](#bootstrap-prompt)
|
|
37
38
|
- [Resume anytime](#resume-anytime)
|
|
@@ -56,6 +57,9 @@ templates — so you always end up with usable output.
|
|
|
56
57
|
|
|
57
58
|
And for those wondering what Payo means, in Gujarati it means foundation — what a building or project is built on.
|
|
58
59
|
|
|
60
|
+
Why I built it, and what I learned along the way:
|
|
61
|
+
[Getting an AI assistant to follow your repo's rules, not just read them](https://uttamgelot.com/notes/why-i-built-payo/).
|
|
62
|
+
|
|
59
63
|
## Why Payo?
|
|
60
64
|
|
|
61
65
|
Every AI coding session starts cold. Your assistant doesn't know you use Drizzle
|
|
@@ -136,6 +140,8 @@ Copilot, Codex, and Antigravity. It plans both
|
|
|
136
140
|
layers together, so each check runs in exactly one place — the agent is told not
|
|
137
141
|
to re-run what a hook already runs.
|
|
138
142
|
|
|
143
|
+

|
|
144
|
+
|
|
139
145
|
**6. One canonical copy, every tool.** One `SKILL.md` per topic under
|
|
140
146
|
`.agents/skills/`, with Claude Code and Windsurf pointed at it by symlink. No
|
|
141
147
|
per-tool duplicates to edit twice and no copies to drift apart — which is
|
|
@@ -220,6 +226,9 @@ guidance files straight into your repo.
|
|
|
220
226
|
answers produced — every one preselected — so you can uncheck anything you
|
|
221
227
|
don't want a file for before a single one is written. Skipped when there's
|
|
222
228
|
only zero or one skill to decide between.
|
|
229
|
+
|
|
230
|
+
<img src="https://raw.githubusercontent.com/uttam-gelot/payo/main/assets/skill-list.png" alt="Payo shows the stack summary, then the list of skills it will generate, each preselected" width="720" />
|
|
231
|
+
|
|
223
232
|
6. **Payo generates the guidance.** It writes one universal layout that every
|
|
224
233
|
skills-compatible tool reads (see the table below). With an agent CLI
|
|
225
234
|
installed, skills are generated in parallel by the AI; otherwise solid
|
|
@@ -331,22 +340,8 @@ agent uses it at that point to check your pending change against the relevant pr
|
|
|
331
340
|
skills and flag anything that conflicts. It reads the diff and reports — it never runs
|
|
332
341
|
your tests, linter or formatter, because your git hook already does.
|
|
333
342
|
|
|
334
|
-
Payo can also add a **pr-review** skill
|
|
335
|
-
|
|
336
|
-
current branch, or a PR / MR by number. Before it reads any code, it asks what to focus
|
|
337
|
-
on (bugs, security, code quality, tests), which level of findings you want, and whether
|
|
338
|
-
to check the change against your project skills. In Claude Code that is one
|
|
339
|
-
`AskUserQuestion` prompt. It checks each finding against the surrounding code before
|
|
340
|
-
reporting it, labels it the [Conventional Comments](https://conventionalcomments.org/)
|
|
341
|
-
way (`issue (blocking):`, `suggestion:`, `nitpick:`, and so on), and puts the worst
|
|
342
|
-
first. Then you pick what happens next: stop, let it fix the issues in your own change,
|
|
343
|
-
post one review on the PR / MR, or get inline comments to paste yourself.
|
|
344
|
-
|
|
345
|
-
It fetches a PR with plain git into a temporary ref, and never checks it out or runs its
|
|
346
|
-
code. It needs `gh` or `glab` only to read a PR description or to post, and it never
|
|
347
|
-
approves a PR or requests changes. Claude Code's `/code-review` and Codex's `/review`
|
|
348
|
-
are built-in commands, so typing them runs those instead. Ask in words ("review PR 42")
|
|
349
|
-
or type `/pr-review`.
|
|
343
|
+
Payo can also add a **pr-review** skill for code and PR reviews. See
|
|
344
|
+
[Code and PR reviews](#code-and-pr-reviews).
|
|
350
345
|
|
|
351
346
|
### Guardrails enforced by hooks, not just prose
|
|
352
347
|
|
|
@@ -427,6 +422,40 @@ your stack needs. Dimensions covered:
|
|
|
427
422
|
the team defaults: what you review, which focus areas to recommend, how findings
|
|
428
423
|
read, and a team checklist every review checks
|
|
429
424
|
|
|
425
|
+
## Code and PR reviews
|
|
426
|
+
|
|
427
|
+
Ask a plain assistant to "review this PR" and you get a generic review. It doesn't
|
|
428
|
+
know your conventions, it re-runs checks your git hooks already run, and it reports
|
|
429
|
+
findings it never verified. Payo can add a **pr-review** skill
|
|
430
|
+
(`.agents/skills/pr-review/SKILL.md`) that fixes that. One question turns it on, and
|
|
431
|
+
one gate sets your team's defaults: what you review, which focus areas to recommend,
|
|
432
|
+
how findings read, and an optional team checklist.
|
|
433
|
+
|
|
434
|
+

|
|
435
|
+
|
|
436
|
+
**What it reviews.** Your local changes, the current branch, or a GitHub / GitLab PR
|
|
437
|
+
or MR by number. It never assumes `main` is the base.
|
|
438
|
+
|
|
439
|
+
**It asks before it reads any code.** What to focus on (bugs, security, code quality,
|
|
440
|
+
tests), which level of findings you want, and whether to check the change against
|
|
441
|
+
your project skills. In Claude Code that is one `AskUserQuestion` prompt.
|
|
442
|
+
|
|
443
|
+
**Verified, labeled, ordered.** It checks each finding against the surrounding code
|
|
444
|
+
before reporting it, labels it the [Conventional Comments](https://conventionalcomments.org/)
|
|
445
|
+
way (`issue (blocking):`, `suggestion:`, `nitpick:`, and so on), and puts the worst
|
|
446
|
+
first.
|
|
447
|
+
|
|
448
|
+
**You pick what happens next.** Stop, let it fix the issues in your own change, post
|
|
449
|
+
one review on the PR / MR after a confirm, or get inline comments to paste yourself.
|
|
450
|
+
|
|
451
|
+
**Safe with untrusted code.** It fetches a PR with plain git into a temporary ref, and
|
|
452
|
+
never checks it out or runs its code. It needs `gh` or `glab` only to read a PR
|
|
453
|
+
description or to post, and it never approves a PR or requests changes.
|
|
454
|
+
|
|
455
|
+
**How to start it.** Ask in words ("review PR 42") or type `/pr-review`. Claude Code's
|
|
456
|
+
`/code-review` and Codex's `/review` are built-in commands, so typing them runs those
|
|
457
|
+
instead.
|
|
458
|
+
|
|
430
459
|
## AI vs. template generation
|
|
431
460
|
|
|
432
461
|
Payo always leaves you with output. When the chosen assistant's headless CLI
|
|
@@ -468,20 +497,16 @@ generating what's missing. Finished runs clean the directory up automatically.
|
|
|
468
497
|
|
|
469
498
|
Payo works today, but it's still early. Here's where it's headed:
|
|
470
499
|
|
|
471
|
-
- **
|
|
472
|
-
|
|
473
|
-
|
|
474
|
-
|
|
475
|
-
|
|
476
|
-
|
|
477
|
-
|
|
478
|
-
|
|
479
|
-
|
|
480
|
-
|
|
481
|
-
supported — so the guidance fits more of the ecosystem out of the box.
|
|
482
|
-
- **A smoother flow.** Faster, clearer prompts; better defaults and grouping;
|
|
483
|
-
tighter summaries and confirmations; and a generation step that's quicker and
|
|
484
|
-
more transparent about what it's doing.
|
|
500
|
+
- **Add one skill without the full interview.** Today Payo asks every question,
|
|
501
|
+
even if you only want pr-review. Next, you name the skill and answer only its
|
|
502
|
+
questions.
|
|
503
|
+
- **More languages, libraries and frameworks**, with recommendations from people
|
|
504
|
+
who work in those stacks every day.
|
|
505
|
+
- **An audit flow.** Payo compares your skills with the current codebase and
|
|
506
|
+
shows where they no longer match.
|
|
507
|
+
- **A better flow for existing projects**, including per-package skills for
|
|
508
|
+
monorepos where each workspace wants its own conventions, and workspace
|
|
509
|
+
enumeration for the remaining ecosystems (Python, PHP, .NET, Ruby).
|
|
485
510
|
|
|
486
511
|
Have a stack you want supported or an idea to make the flow better?
|
|
487
512
|
[Open an issue](https://github.com/uttam-gelot/payo/issues) or see
|