Using Data CLI

Using agent-data-cli

Overview

agent-data-cli is a local information center for multi-source content.

Use it for:

• news
• social media
• financial data
• RSS feeds
• other sources that fit the source/channel/content model

This skill is operational, not explanatory. When the user's goal is clear, execute the CLI flow directly instead of only suggesting commands.

When to Use

Use this skill when the user wants to:

• inspect available sources or channels
• discover remote content before subscribing
• subscribe to channels for ongoing tracking
• sync subscribed channels into the local database
• query locally stored content
• run explicit remote interactions on content refs

Do not use this skill for:

• implementing or redesigning a source
• changing the command surface or core protocol
• free-form scraping outside the agent-data-cli model

Install And Repo Setup

If agent-data-cli is not present locally, install it first:

git clone https://github.com/severinzhong/agent-data-cli
cd agent-data-cli
uv sync

Then:

1. Load the bundled skills from this repository's skills/ directory.
2. Use the repo root that contains pyproject.toml, cli/, sources/, and store/.
3. Execute commands from the repo root.

Always prefer:

uv run -m adc ...

Operating Rules

• Translate natural language into the smallest correct CLI flow.
• Respect the command semantics exactly.
• Do not invent fallback behavior when a capability is unsupported.
• Do not hide remote side effects behind search or update.
• Do not turn content query into a remote search.
• Do not auto-subscribe unless the user's goal implies ongoing tracking.
• For content interact, require an explicit source and explicit refs.

Read references/command-semantics.md before using a command family you have not touched in the current session.

Read references/task-patterns.md when the user request is ambiguous and needs stable intent-to-command mapping.

Read references/result-reporting.md before reporting back after a multi-step run.

Usage Tips

Configure Proxy

When a source needs a proxy, configure it explicitly instead of depending on shell environment:

uv run -m adc config source set <source> proxy_url http://127.0.0.1:7890

If multiple sources should share one proxy, set the CLI-level default:

uv run -m adc config cli set proxy_url http://127.0.0.1:7890

Inspect current source config:

uv run -m adc config source list <source>

Use --jsonl with jq or awk

For machine filtering, prefer --jsonl and pipe to shell tools:

uv run -m adc content query --source cryptocompare --channel BTC --limit 30 --jsonl | jq '.title'
uv run -m adc content query --source cryptocompare --channel BTC --limit 30 --jsonl | jq 'select(.channel_key=="BTC")'
uv run -m adc content query --source cryptocompare --channel BTC --limit 30 --jsonl | awk -F'"' '/"channel_key": "BTC"/ {print $0}'

The same pattern works for remote discovery:

uv run -m adc channel search --source cryptocompare --query BTC --limit 5 --jsonl | jq '.channel_key'

Save Output with > and >>

Use > to overwrite a file and >> to append:

uv run -m adc content query --source cryptocompare --channel BTC --limit 100 --jsonl > btc.jsonl
uv run -m adc content query --source cryptocompare --channel ETH --limit 100 --jsonl >> btc.jsonl
uv run -m adc channel search --source cryptocompare --query BTC --limit 20 --jsonl > channels.jsonl

This is useful when you want to:

• keep a snapshot for later analysis
• feed results into jq, awk, or other CLI tools
• accumulate multiple command outputs into one JSONL file

Execution Flow

1. Classify the request as discovery, subscription, sync, local query, or remote interact.
2. Check whether the task should stay local or requires remote execution.
3. If needed, inspect source capability and config state before executing the main command.
4. Run the shortest correct command chain.
5. Report what was done, what was found, and what next action is now available.

Hard Boundaries

• channel search is remote channel discovery only.
• content search is remote content discovery only and does not write to the database.
• content update is the only remote sync path that writes to the database.
• content query is local-only and never triggers remote work.
• content interact is explicit remote side effect execution only.

If a task does not fit these boundaries, say so directly instead of approximating.