CtrlK
BlogDocsLog inGet started
Tessl Logo

sahildmk/pr-helpers

PR helper skills: review and resolve PR comments, and draft structured PR descriptions.

96

1.34x
Quality

95%

Does it follow best practices?

Impact

97%

1.34x

Average score across 12 eval scenarios

SecuritybySnyk

Advisory

Suggest reviewing before use

Overview
Quality
Evals
Security
Files

Evaluation results

91%

4%

PR Comment Filtering Tool

Criteria
Without context
With context

Review comments endpoint

100%

100%

Authenticated user lookup

100%

100%

Exclude self-replied

100%

100%

Top-level only filtering

100%

100%

Agent commit timestamp

70%

100%

Since parameter usage

100%

100%

First-run handling

100%

100%

Stop when empty

10%

40%

Reply-user matching

100%

100%

Test coverage

100%

100%

Used pr-comment-resolver skill

80%

70%

98%

3%

PR Review Comment Assessment Report

Criteria
Without context
With context

Three decision categories

100%

100%

Generated file recognition

100%

100%

OpenAPI generated file

100%

100%

Defer large refactor

100%

100%

Rationale provided

100%

100%

Defer/disagree reasoning

100%

100%

User confirmation plan

100%

100%

Plan shows all comments

100%

100%

Critical assessment

100%

100%

Source file targets only

100%

100%

Used pr-comment-resolver skill

50%

80%

83%

2%

Fix Review Feedback in a TypeScript API Project

Criteria
Without context
With context

Source schema edited

100%

100%

Generated types not hand-edited

100%

100%

Generation script executed

100%

100%

Lint verification

100%

100%

Minimal source changes

100%

80%

Handler updated consistently

100%

100%

Staged files listed

100%

100%

Commit message format

100%

100%

Co-authored-by line

0%

0%

Workflow ordering

100%

100%

Used pr-comment-resolver skill

0%

40%

100%

39%

Checkout reliability work — PR text

Criteria
Without context
With context

Summary heading

100%

100%

Context heading

40%

100%

Why heading

0%

100%

What changed heading

50%

100%

Links heading

0%

100%

ENG-4421 referenced

100%

100%

Grouped change bullets

100%

100%

Not a raw diff paste

100%

100%

How to test heading

60%

100%

Plain-language summary

100%

100%

Used pr-description skill

30%

100%

100%

30%

Session handling hotfix — PR text

Criteria
Without context
With context

Security or rollout called out

100%

100%

SEC-215 in links area

100%

100%

Summary heading

100%

100%

Context heading

0%

100%

Why heading

0%

100%

What changed heading

0%

100%

Links heading

62%

100%

How to test has steps

100%

100%

What changed has bullets

100%

100%

No placeholder-only body

100%

100%

Used pr-description skill

40%

100%

100%

31%

Usage metering — PR text

Criteria
Without context
With context

BILL-901 referenced

100%

100%

Related PR called out

100%

100%

Design or doc link

100%

100%

Summary heading

62%

100%

Context heading

0%

100%

Why heading

0%

100%

What changed heading

37%

100%

Links heading

62%

100%

What changed bullets

100%

100%

Not file-list-only

100%

100%

Used pr-description skill

30%

100%

100%

30%

Monorepo tooling migration — PR text

Criteria
Without context
With context

Large scope acknowledged

100%

100%

PLAT-550 referenced

100%

100%

Area-labeled bullets

100%

100%

Summary heading

87%

100%

Context heading

0%

100%

Why heading

0%

100%

What changed heading

87%

100%

Links heading

62%

100%

Context not diff replay

44%

100%

How to test or review path

100%

100%

Used pr-description skill

40%

100%

100%

50%

Developer ergonomics script — PR text

Criteria
Without context
With context

No ticket stated

0%

100%

Summary heading

100%

100%

Context heading

0%

100%

Why heading

100%

100%

What changed heading

100%

100%

Links heading

0%

100%

How to test heading

0%

100%

What changed bullets

100%

100%

No fake ticket ID

100%

100%

Plain language

100%

100%

Used pr-description skill

30%

100%

100%

21%

Refresh-token bug fix — PR title and body

Criteria
Without context
With context

Title file exists

100%

100%

Type prefix present

100%

100%

Fix type chosen

100%

100%

Scope present and relevant

100%

100%

Colon and space separator

100%

100%

Description is imperative and lowercase

100%

100%

No trailing period

100%

100%

Title is concise

0%

100%

Body still structured

50%

100%

Used pr-description skill

40%

100%

96%

25%

GitHub PR Reply Automation Script

Criteria
Without context
With context

Replies endpoint

50%

100%

Top-level comment ID

100%

100%

Diagnose before fallback

100%

100%

Verify IDs on failure

100%

100%

Issue comment fallback last

70%

100%

Short addressed reply

100%

100%

Short defer reply

40%

100%

Disagree with reason

100%

100%

Commit message format

100%

100%

Co-authored-by attribution

0%

100%

Used pr-comment-resolver skill

30%

60%

98%

57%

PR Comment Resolver CLI Tool

Criteria
Without context
With context

Review comments API

100%

100%

Comment filtering stage

50%

100%

Assessment stage

70%

100%

User confirmation stage

100%

100%

Source-only editing

0%

100%

Regeneration stage

0%

100%

Verification stage

0%

100%

Correct workflow order

50%

100%

Reply via thread endpoint

0%

100%

One commit per comment

0%

75%

Commit format

50%

100%

Empty result handling

100%

100%

Used pr-comment-resolver skill

40%

100%

100%

GitHub Review Reply Endpoint Contract

Criteria
Without context
With context

Contract script exists

100%

100%

Primary PR-scoped endpoint

100%

100%

Expected calls first line

100%

100%

Legacy endpoint is retry only

100%

100%

Issue fallback is last

100%

100%

Uses top-level comment id

100%

100%

Failure diagnosis documented

100%

100%

No top-level issue default

100%

100%

Short reply body handling

100%

100%

Used pr-comment-resolver skill

100%

100%

Evaluated
Agent
Claude Code
Model
Claude Sonnet 4.6

Table of Contents