• enhancement
• good first issue (optional — likely small but touches multiple dispatchers)

Problem / Motivation

Clawdbot already shows helpful response prefixes (model, provider, thinking level), but users still have to run a separate /status command to see context usage. That breaks flow and makes it easy to accidentally hit the context limit. Other tools (Codex/Claude Code) show a live context indicator in-line, which helps users decide when to compact or start a fresh session.

Goal: expose context window usage as a template variable so it can appear in every response header with no extra commands.

Current Behavior (Before)

Response prefix config:

{
  "messages": {
    "responsePrefix": "[{model} | think:{think}]"
  }
}

Output example:

[claude-opus-4-5 | think:high] Here’s your response...

To see context usage, users must run /status:

📚 Context: 45000/200000 (23%) · 🧹 Compactions: 1

Desired Behavior (After)

Response prefix config:

{
  "messages": {
    "responsePrefix": "[{model} | think:{think} | {context}]"
  }
}

Output example:

[claude-opus-4-5 | think:high | 23%] Here’s your response...

Optional richer template:

{
  "messages": {
    "responsePrefix": "[{model} | {contextUsed}/{contextMax} | {compactions}x]"
  }
}

Proposed Solution

Add new template variables to the response prefix interpolation system:

• {context} / {contextPercent} → percentage of context used (e.g., 23%)
• {contextUsed} → tokens used (e.g., 45000)
• {contextMax} → max tokens (e.g., 200000)
• {compactions} → number of compactions (e.g., 1)

Technical Implementation Notes

Files of interest

• dist/auto-reply/reply/response-prefix-template.js
• Dispatcher(s): dist/telegram/bot-message-dispatch.js (and similar in discord/slack)

1) Extend prefix context type
Add optional fields:

interface ResponsePrefixContext {
  model?: string;
  modelFull?: string;
  provider?: string;
  thinkingLevel?: string;
  identityName?: string;

  // new
  contextPercent?: number;
  contextUsed?: number;
  contextMax?: number;
  compactions?: number;
}

2) Add template variable cases

case "context":
case "contextpercent":
  return context.contextPercent != null ? `${context.contextPercent}%` : match;
case "contextused":
  return context.contextUsed != null ? String(context.contextUsed) : match;
case "contextmax":
  return context.contextMax != null ? String(context.contextMax) : match;
case "compactions":
  return context.compactions != null ? String(context.compactions) : match;

3) Populate context stats
In each dispatcher, extend prefixContext when session stats are available:

prefixContext.contextPercent = Math.round((session.totalTokens / session.maxTokens) * 100);
prefixContext.contextUsed = session.totalTokens;
prefixContext.contextMax = session.maxTokens;
prefixContext.compactions = session.compactionCount;

If context stats are computed later than model selection, consider adding a small hook (e.g., onContextUpdate) or deferring template resolution until stats are known.

4) Docs / schema
Document new variables in config schema and help text for messages.responsePrefix.

Edge Cases / Considerations

• If stats are unavailable, leave the {context*} placeholders untouched (current behavior for unresolved vars).
• First message or no session data: choose whether to show 0% or omit.
• Performance: compute once per response; no heavy recalculation required.

Why This Is Valuable

• Proactive context awareness without extra commands
• Aligns Clawdbot UX with Codex/Claude Code
• Minimal config change for users, backwards compatible by default

Acceptance Criteria

• New template variables {context}, {contextPercent}, {contextUsed}, {contextMax}, {compactions} are supported.
• Response prefix renders context percentage when available.
• Unresolved variables remain literal (no breaking changes).
• Docs updated to mention new variables.