claude_profile 1.4.1

Claude Code account credential management and token status
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

# Algorithm: Subprocess Effort Resolution

### Purpose

Select the `--effort` flag value for isolated subprocess invocations, defaulting to `low` for all effort-capable models and no flag for Haiku or `keep`.

### Entry Point

`src/usage/subprocess.rs` — `resolve_effort(resolved_model, effort_param)`

### Decision Table

| `effort_param` | `resolved_model` | `--effort` injected |
|---|---|---|
| `"low"` | any | `low` |
| `"normal"` | any | `normal` |
| `"high"` | any | `high` |
| `"max"` | any | `max` |
| `"auto"` | `Specific("claude-opus-4-6")` | `low` |
| `"auto"` | `Specific("claude-sonnet-4-6")` | `low` |
| `"auto"` | `Specific("claude-haiku-4-5-20251001")` | *(none)* — Haiku has no extended thinking |
| `"auto"` | `KeepCurrent` | *(none)* — unknown model; avoid injecting effort flag |
| `"auto"` | other Specific | `low` (conservative default) |

### Rationale for `auto` → `low`

Isolated subprocesses run `["--print", "."]` keep-alive prompts. Low effort prevents extended thinking which would cause timeout without adding value.

### `imodel::keep` + `effort::auto` Interaction

When `imodel::keep`, no model is known at dispatch time → `effort::auto` resolves to no flag. This is the safest behavior: avoids injecting an effort flag for an unknown model.

### `imodel::haiku` + `effort::auto` Interaction

Haiku has no extended thinking support → no `--effort` flag. Explicit `effort::low/normal/high/max` with `imodel::haiku` pass through to the subprocess (Claude CLI may ignore or reject).

### Cross-References

| File | Relationship |
|------|-------------|
| [feature/026_subprocess_model_effort.md](../feature/026_subprocess_model_effort.md) | Full feature spec; `effort::` parameter values (AC-05 through AC-16) |
| [algorithm/001](001_touch_model_selection.md) | `resolved_model` input to this algorithm |