@tractorscorch/clank 1.3.1 → 1.4.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -6,6 +6,49 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/).
6
6
 
7
7
  ---
8
8
 
9
+ ## [1.4.1] — 2026-03-23
10
+
11
+ ### Security
12
+ - **Config get redaction** — `config get` action now redacts sensitive keys (apiKey, token, botToken) before returning to LLM context
13
+ - **Config set protection** — config tool now blocks prototype pollution (`__proto__`, `constructor`, `prototype`)
14
+ - **Rate limit streaming path** — `handleInboundMessageStreaming` now enforced (was bypassing rate limiter)
15
+ - **SSRF private IPs** — web_fetch now blocks RFC 1918 ranges (10.x, 192.168.x, 172.16-31.x) and IPv4-mapped IPv6
16
+ - **STT workspace containment** — speech_to_text tool now uses guardPath() to prevent reading files outside workspace
17
+
18
+ ### Audit Result
19
+ - 0 dependency vulnerabilities
20
+ - 14 PASS, 1 WARN (bash blocklist is defense-in-depth), 0 FAIL
21
+ - Grade: A
22
+
23
+ ---
24
+
25
+ ## [1.4.0] — 2026-03-23
26
+
27
+ ### Added
28
+ - **Telegram streaming** — responses edit in real-time as tokens arrive (800ms interval, respects rate limits)
29
+ - **Telegram image handling** — send photos to the bot, routed to agent with image URL
30
+ - **Telegram document handling** — send files to the bot, saved to temp with sanitized filenames, agent can read them
31
+ - **File share tool** — `share_file` lets the agent send workspace files through channels (workspace containment enforced)
32
+ - **Per-agent voice** — each agent can have its own ElevenLabs voice ID in config
33
+ - **Compact prompt mode** — `compactPrompt: true` strips workspace files for small model context optimization
34
+ - **Thinking control** — `thinking: "off"` suppresses extended reasoning for faster responses
35
+ - **Auto-memory persistence** — "remember X", preference statements, and corrections auto-saved to MEMORY.md
36
+ - **Web UI session history** — loads last 50 messages on connect and session switch
37
+ - **Rate limiting** — 20 messages per minute per session, prevents model flooding
38
+
39
+ ### Improved
40
+ - **Model retry** — one retry with 2s backoff on transient connection failures
41
+ - **Session resume compaction** — auto-compacts on load if over context budget
42
+ - **Memory persistence instruction** — system prompt now encourages the agent to save learnings
43
+
44
+ ### Security
45
+ - Telegram document uploads: filename sanitized (path traversal protection), 10MB size limit
46
+ - File share tool: workspace containment via guardPath
47
+ - Rate limiting prevents denial-of-service via message flooding
48
+ - Per-agent voice IDs read from config only (not from user input)
49
+
50
+ ---
51
+
9
52
  ## [1.3.1] — 2026-03-23
10
53
 
11
54
  ### Fixed