docs: add permissionMode field to agent development docs

Document the permissionMode agent frontmatter field (added in v2.0.43)
with kebab-case values for consistency:
- default: inherit from parent conversation
- ask: always prompt for permission
- accept-edits: auto-accept file edits

Note backwards compatibility with camelCase 'acceptEdits' while
establishing kebab-case as the preferred format to match CLI flags.
This commit is contained in:
Claude
2025-12-12 20:06:00 +00:00
parent 2192c86c20
commit 416b2c53c5

View File

@@ -42,6 +42,7 @@ assistant: "[How assistant should respond and use this agent]"
model: inherit model: inherit
color: blue color: blue
tools: ["Read", "Write", "Grep"] tools: ["Read", "Write", "Grep"]
permissionMode: default
--- ---
You are [agent role description]... You are [agent role description]...
@@ -159,6 +160,41 @@ tools: ["Read", "Write", "Grep", "Bash"]
- Testing: `["Read", "Bash", "Grep"]` - Testing: `["Read", "Bash", "Grep"]`
- Full access: Omit field or use `["*"]` - Full access: Omit field or use `["*"]`
### permissionMode (optional)
Control how permission prompts are handled for the agent.
**Format:** String
**Options:**
- `default` - Inherit permission mode from parent conversation (default behavior)
- `ask` - Always prompt for permission before tool use
- `accept-edits` - Auto-accept file edits without prompting
```yaml
permissionMode: accept-edits
```
**Default:** If omitted, inherits from parent conversation settings
**Backwards compatibility:** `acceptEdits` is also accepted but `accept-edits` (kebab-case) is preferred for consistency with CLI flags.
**Use cases:**
- Agents performing many file operations: Use `accept-edits`
- Security-sensitive agents: Use `ask` to ensure oversight
- General purpose: Omit field to inherit conversation settings
**Example:**
```yaml
---
name: code-generator
description: Use this agent when...
model: inherit
color: green
permissionMode: accept-edits
---
```
## System Prompt Design ## System Prompt Design
The markdown body becomes the agent's system prompt. Write in second person, addressing the agent directly. The markdown body becomes the agent's system prompt. Write in second person, addressing the agent directly.
@@ -355,6 +391,7 @@ Output: [What to provide]
| model | Yes | inherit/sonnet/opus/haiku | inherit | | model | Yes | inherit/sonnet/opus/haiku | inherit |
| color | Yes | Color name | blue | | color | Yes | Color name | blue |
| tools | No | Array of tool names | ["Read", "Grep"] | | tools | No | Array of tool names | ["Read", "Grep"] |
| permissionMode | No | default/ask/accept-edits | accept-edits |
### Best Practices ### Best Practices