Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions docs.json
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,12 @@
]
},
"ecosystem/sdks",
{
"group": "AI",
"pages": [
"ecosystem/ai/agent-skill"
]
},
{
"group": "APIs",
"pages": [
Expand Down
43 changes: 43 additions & 0 deletions ecosystem/ai/agent-skill.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,43 @@
---
title: "Agent skill"
---

## Objective

Install the `ton-blockchain` agent Skill so your AI coding agent follows a docs-first workflow when you work on TON-related tasks.

## About Agent Skills

Agent Skills are an open format for packaging instructions, scripts, and resources that skills-compatible agents can discover and apply.

See [https://agentskills.io/home](https://agentskills.io/home) for an overview and links to the specification.

## Prerequisites

- The agent you use supports “Skills” (a `SKILL.md` file with frontmatter)
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[HIGH] Reader-directed pronoun and decorative quotes in prerequisites

The prerequisite bullet on line 17 uses the phrase “The agent you use”, which addresses the reader directly in the second person. The same line also wraps the term “Skills” in quotation marks used for emphasis rather than as a literal UI string. The style guide’s “Don’t get personal” and quotation-mark rules require neutral phrasing without “you/your” and discourage decorative quotes for emphasized terminology. Rewriting this bullet to use third-person, neutral language and plain capitalization for Skills keeps the meaning while conforming to the house style.

Suggested change
- The agent you use supports “Skills” (a `SKILL.md` file with frontmatter)

Please leave a reaction 👍/👎 to this suggestion to improve future reviews for everyone!


## Install in Codex

Run this in Codex:
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[HIGH] Reader-directed phrasing in Codex instructions

Line 21 reads “Run this in Codex:”, which implicitly addresses the reader and uses conversational phrasing. The style guide for documentation favors neutral, imperative instructions that describe what to do without talking to “you” as a person. Rephrasing this line to refer to the command itself rather than the reader keeps the step clear while aligning with the required tone.

Suggested change
Run this in Codex:
Run the following command in Codex:

Please leave a reaction 👍/👎 to this suggestion to improve future reviews for everyone!


```
$skill-installer --url https://github.com/ton-org/docs/tree/main/skills/ton-blockchain
```
Comment on lines +23 to +25
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[HIGH] Shell prompt token and missing language tag in skill-installer command

The fenced block on lines 23–25 shows the command as $skill-installer ... inside an untyped code fence. Including a $-style prompt in command examples conflicts with the guideline that examples must be copy-pasteable without editing out prompts. The same block also omits a language identifier, whereas other command examples in the file consistently use bash, and the style guide requires specifying a language for all fenced code blocks. Updating this block to remove the prompt token and to label the fence with bash addresses both issues and keeps command examples consistent.

Please leave a reaction 👍/👎 to this suggestion to improve future reviews for everyone!


Then restart Codex to pick up the new skill.

## Install in Claude Code

This installs it as a personal Skill:
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[HIGH] Reader-directed phrasing in Claude Code instructions

The sentence on line 31, “This installs it as a personal Skill:”, uses conversational, reader-targeted phrasing rather than a neutral imperative. The documentation style guide prefers direct instructions that describe the action to take, without framing the text as a commentary for the reader. Rewriting this line into an imperative instruction keeps the meaning and aligns with the neutral, fact-focused tone.

Suggested change
This installs it as a personal Skill:
Install this as a personal Skill:

Please leave a reaction 👍/👎 to this suggestion to improve future reviews for everyone!


```bash
mkdir -p ~/.claude/skills/ton-blockchain && \
curl -fsSL https://raw.githubusercontent.com/ton-org/docs/main/skills/ton-blockchain/SKILL.md \
-o ~/.claude/skills/ton-blockchain/SKILL.md
```

Then restart Claude Code to pick up the new skill.

## Install in other agents

Copy `skills/ton-blockchain/SKILL.md` into your agent’s supported Skill directory (for example a project Skill folder), keeping the directory name `ton-blockchain`.
58 changes: 58 additions & 0 deletions skills/ton-blockchain/SKILL.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,58 @@
---
name: ton-blockchain
description: "Use when working with The Open Network (TON) blockchain, or when the user mentions TON-ecosystem terms the agent may not recognize—such as Tact, FunC, Tolk, Fift, TL-B, TVM, cells, BoC, Jettons, TEPs, TON Connect, workchains, shardchains, or liteservers. Provides a docs-first workflow for fetching and navigating TON documentation accurately, and covers smart contract development, transaction mechanics, wallet standards, token standards, Telegram Mini Apps, and infrastructure."
---

# TON Docs-first workflow

Prefer primary sources from TON Docs.

## 1) Read the orientation page first

Fetch and skim the TON Docs “start here” page to align terminology and the docs’ structure before making assumptions:

```bash
curl -fsSL https://docs.ton.org/start-here.md
```

If the `.md` URL fails, try the HTML page:

```bash
curl -fsSL https://docs.ton.org/start-here
```

## 2) Discover available pages

Fetch `llms.txt` to get the authoritative list of TON Docs pages and then search within it for relevant sections/pages:

```bash
curl -fsSL https://docs.ton.org/llms.txt
```

Practical pattern:

```bash
curl -fsSL https://docs.ton.org/llms.txt | rg -n "<topic>|<keyword>" || true
```

(Use `grep -nE` instead of `rg` if ripgrep isn’t available.)

## 3) Pull only the pages you need

For any relevant page path from `llms.txt`, prefer the Markdown source by appending `.md`:

```bash
curl -fsSL "https://docs.ton.org/<page-path>.md"
```

If that 404s, try the path as-is (some entries may already include `.md` or may not support source rendering):

```bash
curl -fsSL "https://docs.ton.org/<page-path>"
Comment on lines +34 to +51
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[HIGH] Undocumented <ANGLE_CASE> placeholders in commands

The workflow introduces <topic>, <keyword>, and <page-path> placeholders directly inside runnable curl and rg commands (for example in the search pattern on line 35 and in the page fetch commands on lines 45 and 51) without defining what values should be substituted. The style guide requires that angle-bracketed placeholders in commands be explicitly defined on first use so examples remain copy-pasteable and unambiguous. Without short definitions, users and tools may misinterpret these parameters or be unsure how to construct valid queries and paths.

Please leave a reaction 👍/👎 to this suggestion to improve future reviews for everyone!

```

## 4) Execute the task using the docs as ground truth

- Treat TON Docs as the primary reference; reconcile contradictions explicitly.
- When answering, mention which TON Docs pages you consulted (page titles/paths).
- If the user’s code/repo conventions conflict with TON Docs guidance, ask for clarification before proceeding.