一键在 Manus 中运行任何 Skill

amos-graph

星标0

分支0

更新时间2026年4月20日 23:28

Print the amos dependency tree as a nested ASCII graph. Use when the user asks about open tasks, the roadmap, what's next, what's blocked, what depends on X, or wants a structural view of the plan. Returns every node with status and a short description; does NOT resolve `@github:...` references (use amos-show for full content). Trigger on phrases like "what's on my plate", "what's open", "show me the DAG", "what's blocking X", "what depends on Y", "where are we on <umbrella>".

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

tatolab

tatolab/amos

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Output format

Nested tree, indent encodes depth. Each line looks like:

[state=OPEN] #319 GPU capability-based access (sandbox + escalate) — Umbrella — replace runtime-phase checks ...
    ├── [state=CLOSED] #320 Design doc: ... — Write the design doc gating all downstream work ...
    ├── [state=CLOSED] #321 Introduce GpuContext... — Add the two capability types as thin newtype wrappers ...
    └── [state=OPEN] #326 Learning doc: ... — Capture the "why" of the sandbox/escalate pattern ...
        ├── [state=CLOSED] #325 (→ see above)
        ├── [state=OPEN] #369 Polyglot escalate: ... — Follow-up to #325. Extend the polyglot escalate IPC ...
        └── [state=OPEN] #370 xtask: JTD discriminator schemas — Update xtask/src/generate_schemas.rs ...

Legend:

[state=OPEN] / [state=CLOSED] — lifecycle state (from the adapter, e.g. GitHub issue state).
[labels=...] — tracker labels, when present.
#N — GitHub issue number, derived from the canonical @github:<owner>/<repo>#N node name.
Text after the em-dash — the node's description: field (a routing hint).
(→ see above) — back-reference; the subtree was already printed earlier.

When to use

"What's next" / "what's on my plate" / "what's open"
"What's blocking X" / "what does Y depend on"
Finding the next unblocked task under an umbrella
Orientation in a new or unfamiliar amos-tracked project
Spotting broken edges (orphan nodes appear at the top level instead of nested)

When NOT to use

Need the full resolved body of a single node (GitHub issue text, etc.) → use amos-show.
Just want the raw markdown of one plan file → use Read on the file directly (no network, much faster).
Need to post a comment to the tracker → use amos-notify.

Filtering large trees

Output can be long (hundreds of lines on mature projects). To zoom in, pipe through grep with context:

# Show a specific umbrella and its children
"$HOME/.local/bin/amos" graph | grep -A 30 "#319 "

# Find all open tasks
"$HOME/.local/bin/amos" graph | grep "state=OPEN"

# Find orphan nodes (no parent, broken edges)
"$HOME/.local/bin/amos" graph | grep -B 0 "^[A-Z\[]"

同仓库更多 Skills

同仓库

amos-next

tatolab/amos

Pick up and execute the next ready-to-start issue from the focused milestone, end-to-end: announce, gate on user confirmation, branch, do the work, run the tests listed in the issue body, open a PR, report. This is the skill to invoke when the user says "continue", "next task", "what's next", "pick up where we left off", "work on the next issue", "keep going", or similar. It replaces any project-local PROMPT.md protocol.

2026-05-020

amos-file

tatolab/amos

File a new GitHub issue with a well-shaped body, the right milestone, and native dependency relationships, from a short natural-language intent. Use when the user says "file this", "open an issue for X", "create a ticket", "log this as a bug", when discussion surfaces a new task that should be tracked, or after the assistant offered to file something and the user confirmed. Infers milestone and labels; asks in chat whenever confidence is low.

2026-04-270

amos

tatolab/amos

Load when working with amos nodes — markdown files carrying `whoami: amos` frontmatter — or when the user refers to "the plan", "the DAG", "the roadmap", "the tickets", "what's next", or similar work-tracking concepts. Installs the amos conventions (canonical naming, node format, when to reach for which sub-skill) into context so amos-graph, amos-show, amos-create, and amos-notify can be used correctly. Does not run the CLI on its own.

2026-04-210

amos-focus

tatolab/amos

Set the milestone the user is currently working on. `amos next`, `amos blocked`, and `amos graph` scope their output to this milestone once set. Use when the user says "let's work on <milestone>", "focus on <milestone>", "switch to <milestone>", or wants to check which milestone is currently active. Accepts free-form milestone text — the tool resolves it against the adapter's known titles.

2026-04-200

amos-create

tatolab/amos

Scaffold a new amos node with proper frontmatter. Use when the user wants to add a new ticket, task, or plan entry tracked by amos. Produces a markdown file with the canonical `@github:<owner>/<repo>#<N>` name, a description, dependencies, and the adapter block.

2026-04-200

amos-notify

tatolab/amos

Post a message to a node's source system through its adapter. For a `@github:...` node, this creates a comment on the referenced issue. Use when the user wants to share status, results, or context with everyone watching the upstream tracker — not for local scratch notes. The message is visible to anyone with access to the tracker.

2026-04-200

name	amos-graph
description	Print the amos dependency tree as a nested ASCII graph. Use when the user asks about open tasks, the roadmap, what's next, what's blocked, what depends on X, or wants a structural view of the plan. Returns every node with status and a short description; does NOT resolve `@github:...` references (use amos-show for full content). Trigger on phrases like "what's on my plate", "what's open", "show me the DAG", "what's blocking X", "what depends on Y", "where are we on <umbrella>".
allowed-tools	Bash

Run from the project root (the directory containing amos-scanned markdown files):

"$HOME/.local/bin/amos" graph

Output format

Nested tree, indent encodes depth. Each line looks like:

[state=OPEN] #319 GPU capability-based access (sandbox + escalate) — Umbrella — replace runtime-phase checks ...
    ├── [state=CLOSED] #320 Design doc: ... — Write the design doc gating all downstream work ...
    ├── [state=CLOSED] #321 Introduce GpuContext... — Add the two capability types as thin newtype wrappers ...
    └── [state=OPEN] #326 Learning doc: ... — Capture the "why" of the sandbox/escalate pattern ...
        ├── [state=CLOSED] #325 (→ see above)
        ├── [state=OPEN] #369 Polyglot escalate: ... — Follow-up to #325. Extend the polyglot escalate IPC ...
        └── [state=OPEN] #370 xtask: JTD discriminator schemas — Update xtask/src/generate_schemas.rs ...

Legend:

[state=OPEN] / [state=CLOSED] — lifecycle state (from the adapter, e.g. GitHub issue state).
[labels=...] — tracker labels, when present.
#N — GitHub issue number, derived from the canonical @github:<owner>/<repo>#N node name.
Text after the em-dash — the node's description: field (a routing hint).
(→ see above) — back-reference; the subtree was already printed earlier.

When to use

"What's next" / "what's on my plate" / "what's open"
"What's blocking X" / "what does Y depend on"
Finding the next unblocked task under an umbrella
Orientation in a new or unfamiliar amos-tracked project
Spotting broken edges (orphan nodes appear at the top level instead of nested)

When NOT to use

Need the full resolved body of a single node (GitHub issue text, etc.) → use amos-show.
Just want the raw markdown of one plan file → use Read on the file directly (no network, much faster).
Need to post a comment to the tracker → use amos-notify.

Filtering large trees

Output can be long (hundreds of lines on mature projects). To zoom in, pipe through grep with context:

# Show a specific umbrella and its children
"$HOME/.local/bin/amos" graph | grep -A 30 "#319 "

# Find all open tasks
"$HOME/.local/bin/amos" graph | grep "state=OPEN"

# Find orphan nodes (no parent, broken edges)
"$HOME/.local/bin/amos" graph | grep -B 0 "^[A-Z\[]"