@heart-of-gold/toolkit 0.1.13 → 0.1.15

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/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@heart-of-gold/toolkit",
3
- "version": "0.1.13",
3
+ "version": "0.1.15",
4
4
  "type": "module",
5
5
  "description": "Cross-platform installer for Heart of Gold skills — works with Codex, OpenCode, Pi, Claude Code, and more",
6
6
  "bin": {
@@ -37,11 +37,15 @@ Choose the permission mode based on what Claude Code needs to do:
37
37
  `plan` is not the general default for this skill. It is only the safe default for bounded read-only tasks.
38
38
 
39
39
  ## Running a Task
40
- 1. Ask the user which model alias to use (default: `sonnet`) and which permission mode to use when that choice materially affects behavior.
41
- 2. Decide whether Claude should receive the artifact directly or discover it itself:
40
+ 1. Always ask the user which model alias to use (default: `sonnet`) before running Claude Code, unless the user already specified the model.
41
+ 2. Ask for permission mode in the same prompt when the user did not specify it. Default to:
42
+ - `plan` for bounded read-only analysis
43
+ - `acceptEdits` for implementation or refactoring work
44
+ - `default` when Claude needs broader repo exploration without auto-edit authority
45
+ 3. Decide whether Claude should receive the artifact directly or discover it itself:
42
46
  - Pass diffs, logs, or file contents via stdin for safe read-only review
43
47
  - Let Claude inspect the working tree when the task requires tool use
44
- 3. Assemble the command with the appropriate options:
48
+ 4. Assemble the command with the appropriate options:
45
49
  - `-p, --print` for non-interactive output
46
50
  - `--output-format <text|json|stream-json>`
47
51
  - `--model <MODEL>`
@@ -52,12 +56,24 @@ Choose the permission mode based on what Claude Code needs to do:
52
56
  - `--add-dir <DIR>` when Claude must read outside the current working directory
53
57
  - `--name <NAME>` when you want a stable, human-readable session
54
58
  - `"your prompt here"` as the final positional argument
55
- 4. Prefer `--output-format text` for human-readable summaries and `--output-format json` for automation or machine parsing.
56
- 5. In headless automation, prefer explicit permissions:
59
+ 5. Prefer `--output-format text` for human-readable summaries and `--output-format json` for automation or machine parsing.
60
+ 6. In headless automation, prefer explicit permissions:
57
61
  - `plan` for provided artifacts
58
62
  - `acceptEdits` for edit-capable runs
59
63
  - `default` only when prompts are acceptable or when permissions are constrained with `--allowedTools`
60
- 6. Run the command, capture stdout/stderr, and summarize the outcome for the user.
64
+ 7. Run the command, capture stdout/stderr, and summarize the outcome for the user.
65
+
66
+ ## Codex Execution
67
+
68
+ When this skill is used from Codex, the skill cannot bypass Codex sandbox policy on its own. Instead, instruct Codex to choose the correct execution path:
69
+
70
+ - Run inside the sandbox first when the task is clearly local and read-only.
71
+ - If Claude Code needs network access, auth refresh, access outside the writable sandbox, or the in-sandbox run fails or hangs for likely sandbox reasons, rerun via `exec_command` with `sandbox_permissions: "require_escalated"`.
72
+ - Use a concise justification that tells the reviewer why Claude Code needs elevation, for example:
73
+ - "Do you want me to run Claude Code with network access so it can authenticate and complete this review?"
74
+ - "Do you want me to rerun Claude Code outside the sandbox because the in-sandbox headless run appears blocked by permissions?"
75
+ - Prefer a narrow approval request. Ask for elevation only for the Claude invocation that needs it, not unrelated commands.
76
+ - If the task is high-risk, restate the chosen Claude permission mode when asking for elevation.
61
77
 
62
78
  ## Best Practices
63
79
 
@@ -189,5 +205,6 @@ claude -r latest -p \
189
205
  - Stop and report failures whenever `claude --version` or a `claude -p` command exits non-zero.
190
206
  - If Claude Code reports permission issues, choose the correct permission mode or constrain tools explicitly.
191
207
  - If a headless run gets stuck on permissions, either switch to `plan`, use `acceptEdits`, or provide `--allowedTools` / a permission prompt tool.
208
+ - In Codex, if the likely cause is sandboxing or network denial, rerun with reviewer-approved `require_escalated` execution instead of repeatedly retrying the same sandboxed command.
192
209
  - Do not use `bypassPermissions` unless the user explicitly approves it.
193
210
  - If debugging a failing wrapper invocation, fall back to the direct `claude` command first.
package/src/index.ts CHANGED
@@ -7,7 +7,7 @@ import { targetsCommand } from "./commands/targets";
7
7
  const main = defineCommand({
8
8
  meta: {
9
9
  name: "heart-of-gold",
10
- version: "0.1.13",
10
+ version: "0.1.15",
11
11
  description:
12
12
  "Cross-platform installer for Heart of Gold skills — Codex, OpenCode, Pi, Claude Code, and more",
13
13
  },