Twitter Skill (Read-Only)

Step 1: Ensure twitter-cli Is Installed and Authenticated

# Install (requires Python 3.10+)
command -v twitter || uv tool install twitter-cli

# Check authentication
twitter status --yaml >/dev/null && echo "AUTH_OK" || echo "AUTH_NEEDED"

AUTH_OK

AUTH_NEEDED

Authentication methods

twitter whoami

TWITTER_CHROME_PROFILE="Profile 2" twitter feed

export TWITTER_AUTH_TOKEN="<auth_token from browser>"
export TWITTER_CT0="<ct0 from browser>"
twitter whoami

Common auth issues

No Twitter cookies found

Cookie expired (401/403)

Unable to get key for cookie decryption

security unlock-keychain ~/Library/Keychains/login.keychain-db

Step 2: Identify What the User Needs

references/commands.md

twitter status

--yaml

twitter whoami

--json

--yaml

twitter feed

-t following

--max N

--full-text

--filter

twitter search "QUERY"

-t Latest

--from USER

--since DATE

--lang

--exclude retweets

--has links

--max N

twitter bookmarks

--max N

--full-text

twitter tweet TWEET_ID_OR_URL

--full-text

twitter show N

--full-text

--json

twitter article ID_OR_URL

--markdown

--output FILE

--json

twitter list LIST_ID

--full-text

--max N

twitter user USERNAME

--json

twitter user-posts USERNAME

--max N

--full-text

-o FILE

twitter likes USERNAME

twitter followers USERNAME

--max N

twitter following USERNAME

--max N

Step 3: Execute the Command

General pattern

# Use --yaml for structured output, -c for compact (token-efficient)
twitter feed --yaml --max 20
twitter -c feed --max 10              # Compact: ~80% fewer tokens than --json

# Searching for financial topics
twitter search "$AAPL earnings" -t Latest --max 10 --full-text --yaml
twitter search "Fed rate decision" --since 2026-03-01 --max 20 --yaml
twitter search "market crash" --exclude retweets --has links --max 15

Key rules

1. Check auth first — run

   twitter status --yaml

   before any other command
2. Use

   --yaml

   or

   --json

   for structured output when processing data programmatically
3. Use

   -c

   /

   --compact

   when token efficiency matters (LLM context)
4. Use

   --full-text

   when the user wants complete tweet content (text is truncated by default in rich tables)
5. Use

   --max N

   to limit results — start with 10-20 unless the user asks for more
6. For search, use filters —

   --from

   ,

   --since

   ,

   --lang

   ,

   --exclude retweets

   ,

   --has links

7. Tweet IDs vs URLs — both work:

   twitter tweet 1234567890

   or

   twitter tweet https://x.com/user/status/1234567890

8. Use

   twitter show N

   to reference tweet #N from the most recent list output
9. NEVER execute write operations — this skill is read-only; do not post, like, retweet, reply, quote, follow, or delete

Output flags (all read commands)

--json

--yaml

--full-text

-c

--compact

-o FILE

--output FILE

--input FILE

Structured output envelope

--yaml

--json

references/schema.md

ok: true
schema_version: "1"
data: ...          # tweet/user lists or single objects

ok: false
schema_version: "1"
error:
  code: api_error
  message: "..."

Step 4: Present the Results

1. Summarize key content — highlight the most relevant tweets for the user's financial research
2. Include attribution — show @username, tweet text, and engagement metrics (likes, retweets)
3. Provide tweet IDs/URLs when the user might want to read the full thread
4. For search results, group by relevance and highlight key themes, sentiment, or market signals
5. For user profiles, present follower count, bio, and notable recent activity
6. Flag sentiment — note bullish/bearish sentiment, consensus vs contrarian views
7. Treat cookies as secrets — never echo cookie values to stdout

Step 5: Diagnostics

twitter doctor

-v

Error Reference

No Twitter cookies found

Proxy Configuration

export TWITTER_PROXY=http://127.0.0.1:7890
# or
export TWITTER_PROXY=socks5://127.0.0.1:1080

Reference Files

• references/commands.md

  — Complete read command reference with all flags, research workflows, and usage examples

• references/schema.md

  — Structured output schema documentation