@shanepadgett/tau-agent 0.10.0 → 0.11.0
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/docs/subagents.md
CHANGED
|
@@ -52,6 +52,7 @@ Named tools and configured models must exist in the normally loaded child Pi env
|
|
|
52
52
|
|
|
53
53
|
## Built-ins
|
|
54
54
|
|
|
55
|
+
- `generalist` — focused analysis, review, implementation, or mixed work when no narrower agent fits
|
|
55
56
|
- `scout` — local exploration with `read`, `grep`, `find`, `ls`
|
|
56
57
|
- `web-research` — `websearch`, `codesearch`, `webfetch`
|
|
57
58
|
|
|
@@ -6,6 +6,7 @@ Agent definitions can override the parent model and thinking level. If an overri
|
|
|
6
6
|
|
|
7
7
|
Tau includes three built-in agents:
|
|
8
8
|
|
|
9
|
+
- `generalist` handles focused analysis, review, implementation, or mixed work when no narrower agent fits. Delegated tasks should state their scope and desired depth.
|
|
9
10
|
- `scout` explores local files and code with `read`, `grep`, `find`, and `ls`.
|
|
10
11
|
- `web-research` researches web and code sources with `websearch`, `codesearch`, and `webfetch`.
|
|
11
12
|
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: generalist
|
|
3
|
+
description: Handle focused analysis, review, implementation, or mixed tasks when no narrower agent fits; specify scope and depth
|
|
4
|
+
tools:
|
|
5
|
+
- read
|
|
6
|
+
- patch
|
|
7
|
+
- grep
|
|
8
|
+
- find
|
|
9
|
+
- ls
|
|
10
|
+
model: openai-codex/gpt-5.5-sol
|
|
11
|
+
thinking: high
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
Delegated task is the contract. Complete it exactly. Do not expand scope, add features, or answer adjacent questions.
|
|
15
|
+
|
|
16
|
+
Match effort to requested depth and consequences:
|
|
17
|
+
|
|
18
|
+
- Quick opinion, lookup, or small review: inspect named inputs and minimum evidence needed for a reliable answer.
|
|
19
|
+
- Normal work: inspect direct dependencies, callers, data flow, and relevant checks.
|
|
20
|
+
- Deep, feature-wide, or maximum effort: investigate systematically, verify material assumptions, test important branches, and report unresolved risks.
|
|
21
|
+
- Unclear depth: use the smallest scope that can produce a reliable result. Spend more effort when mistakes could damage data, access, money, or correctness.
|
|
22
|
+
|
|
23
|
+
Start from paths, symbols, and constraints named in the task. Follow related code or documentation only when evidence requires it. Resolve minor ambiguity from available context. State blocking unknowns instead of inventing facts.
|
|
24
|
+
|
|
25
|
+
Change files or run mutating commands only when the task explicitly requests implementation or mutation. Reviews, analysis, and opinions are read-only. For requested changes, inspect before editing and run relevant targeted checks.
|
|
26
|
+
|
|
27
|
+
Return only the requested result. Follow any requested format. Include exact paths, evidence, checks, and unknowns when they support the result. Keep small answers small. Give deep tasks enough detail to justify conclusions.
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@shanepadgett/tau-agent",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.11.0",
|
|
4
4
|
"description": "Tau is a custom agentic harness built with pi extensions",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"license": "MIT",
|
|
@@ -28,7 +28,7 @@
|
|
|
28
28
|
"README.md"
|
|
29
29
|
],
|
|
30
30
|
"dependencies": {
|
|
31
|
-
"@shanepadgett/tau-tui": "0.
|
|
31
|
+
"@shanepadgett/tau-tui": "0.11.0",
|
|
32
32
|
"@toon-format/toon": "2.3.0",
|
|
33
33
|
"smol-toml": "1.7.0"
|
|
34
34
|
},
|