tesseract-agent 0.11.34 → 0.11.35
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.
|
@@ -2,12 +2,12 @@
|
|
|
2
2
|
name: continue-from-another-agent
|
|
3
3
|
description: Pick up work that another agent, framework or conversation started — a conversation moved here from another framework, an earlier Tesseract conversation, a session in Claude Code, Codex, Gemini CLI, Cursor or another agent's own app, or a hand-off note — by finding and reading what was done, rebuilding the state, checking it against the project as it is now, and continuing from the exact stopping point. Also for writing a hand-off for the next agent, and for bringing another agent's memories, skills and connectors into Tesseract.
|
|
4
4
|
license: MIT
|
|
5
|
-
version: 1.1.
|
|
5
|
+
version: 1.1.1
|
|
6
6
|
metadata:
|
|
7
7
|
tesseract.summary: Find what another agent or conversation did, verify it against the project now, and carry on from there.
|
|
8
8
|
tesseract.category: software
|
|
9
9
|
tesseract.examples: "continue what I was doing in Codex yesterday | pick up where my last Claude Code session stopped | write a handoff so another agent can finish this | कल दूसरी चैट में जो काम शुरू किया था, उसे आगे बढ़ाओ | أكمل المهمة التي بدأها وكيل آخر"
|
|
10
|
-
tesseract.tools: sessions_search, sessions_read, memory_read, memory_write, app_extensions
|
|
10
|
+
tesseract.tools: sessions_search, sessions_read, memory_read, memory_write, app_extensions, crew_handoff
|
|
11
11
|
---
|
|
12
12
|
|
|
13
13
|
# Continue from another agent
|
|
@@ -100,7 +100,11 @@ Ask one focused question only when the stopping point or the intended next step
|
|
|
100
100
|
## Writing a hand-off for the next agent
|
|
101
101
|
|
|
102
102
|
When the person asks for one, or before they move the work elsewhere: write it in the conversation, or in
|
|
103
|
-
a file they choose (not committed unless asked).
|
|
103
|
+
a file they choose (not committed unless asked). Start from `crew_handoff` (reached with `tool_call`): it
|
|
104
|
+
drafts one from what Tesseract recorded of a session's work, not from what was said — what it was asked,
|
|
105
|
+
the files it changed and its commits, the commands it ran and how each ended, its plan, its leases and what
|
|
106
|
+
it waits on — and `keep` writes it to a Markdown file. Another session of this project can draft one for
|
|
107
|
+
any session here too. Then add what only you know. Include the five points of step 2, the commands to verify
|
|
104
108
|
the current state, the decisions and their reasons, and anything still running (jobs, watches, scheduled
|
|
105
109
|
tasks — with their ids). Write it for a reader with no history: absolute paths, exact commands, no "as
|
|
106
110
|
discussed". Moving this conversation itself to another framework is the person's choice in its framework
|