Brain Entry

Brain Entry

Create a capture in Open Brain from any instruction. The argument tells you what content to generate and send.

Examples

/brain-entry summarize everything we have done so far in this session and send it to my open-brain
/brain-entry log a decision: we chose Cloudflare Email Routing for email ingestion because it adds zero infrastructure
/brain-entry capture this idea: add a weekly email digest that summarizes all captures from the past 7 days
/brain-entry record that I had a productive call with Butch about the CFA merger timeline

Implementation

Step 1: Generate the content

Based on the user's instruction in the arguments:

• If the instruction says "summarize this session" or similar — review the full conversation context and produce a clear, structured summary of what was accomplished, decisions made, and outcomes.
• If the instruction describes a decision, idea, observation, task, or other capture — write it up clearly and concisely.
• If the instruction references specific context from the conversation — include the relevant details.

Write the content in first person (Troy's voice) since this is his personal knowledge base. Be substantive and specific — include names, dates, technical details, and outcomes. No filler.

Step 2: Classify the capture

Determine the best capture_type from: decision, idea, observation, task, win, blocker, question, reflection.

Determine the best brain_view from: career, personal, technical, work-internal, client.

Use your judgment based on the content. When in doubt, use observation and personal.

Step 3: POST to the captures API

Use the Bash tool with curl and a heredoc to POST to the captures API. Always use curl, not Python urllib — Cloudflare blocks Python's default user-agent with 403.

Use a heredoc for the JSON body to avoid shell escaping issues with quotes and newlines in the content. Bound the call with --max-time/--connect-timeout so a hung connection cannot block indefinitely, and capture the HTTP status alongside the body so success is verified before confirming anything to the user:

RAW=$(curl -s -w '\n%{http_code}' --max-time 60 --connect-timeout 10 \
  -X POST "https://brain.troy-davis.com/api/v1/captures" \
  -H "Content-Type: application/json" \
  -H "X-Open-Brain-Caller: claude-code" \
  -d @- <<'ENDJSON'
{
  "content": "<generated content with \n for newlines>",
  "capture_type": "<classified type>",
  "brain_view": "<classified view>",
  "source": "api",
  "metadata": {
    "source_metadata": {
      "origin": "claude-code-skill",
      "session_context": "<brief description of what prompted this>"
    }
  }
}
ENDJSON
)
CURL_EXIT=$?
HTTP_CODE=$(echo "$RAW" | tail -n1)
BODY=$(echo "$RAW" | sed '$d')

if [ "$CURL_EXIT" -ne 0 ] || [ "$HTTP_CODE" -ge 400 ] || [ -z "$BODY" ]; then
  echo "Capture failed: curl_exit=$CURL_EXIT http=$HTTP_CODE"
  echo "$BODY"
  exit 1
fi
echo "$BODY"

Important:

• Use heredoc (<<'ENDJSON') to pass the JSON body — avoids shell escaping entirely
• Content max length: 50,000 characters
• The API returns { id, pipeline_status, created_at } on success (201) — do not declare success until CURL_EXIT is 0 and HTTP_CODE is below 400
• Do NOT use Python urllib — Cloudflare returns 403 on Python's default user-agent

Step 4: Confirm to the user

Only after the Step 3 checks pass, parse id, pipeline_status, and created_at from $BODY (e.g. echo "$BODY" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['id'], d['pipeline_status'], d['created_at'])") and display:

Captured in Open Brain:
  ID: <capture_id>
  Type: <capture_type>
  View: <brain_view>
  Pipeline: <status>

On failure (non-zero CURL_EXIT, HTTP_CODE >= 400, or an empty body), show the error and suggest the user check that the homeserver is running.

Error Handling

• If the API returns 4xx/5xx, show the error body
• If the API is unreachable, suggest checking Cloudflare Tunnel / homeserver status
• If the content exceeds 50K chars, truncate with a [truncated] marker
• Never silently fail — always report the outcome

Notes

• The pipeline will automatically classify, embed, and extract entities from the capture
• The capture will be searchable immediately via FTS and after embedding via semantic search
• This skill is the Claude Code equivalent of voice memos (phone) or Slack messages — another input channel into the brain