chitin-moat
Enforce contextual permission boundaries for AI agents based on communication surface. Constrains agent capabilities (exec, file I/O, secrets, messaging) by channel trust level rather than message content, preventing social engineering and prompt injection in group chats. Use when: (1) configuring agent permissions per channel/group, (2) setting up read-only mode for public Discord/Telegram, (3) implementing sovereign/trusted/guarded/observer/silent trust tiers, (4) auditing agent channel permissions, or (5) the user mentions "trust channels", "channel permissions", or "read-only mode."
来源信息
- 仓库
- knownasnaffy/prompthound
- 最近来源活动
- 2026年7月6日 07:03
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 0
- 分支
- 1
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
文件资源管理器
7 个文件正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- chitin-moat
- description
- Enforce contextual permission boundaries for AI agents based on communication surface. Constrains agent capabilities (exec, file I/O, secrets, messaging) by channel trust level rather than message content, preventing social engineering and prompt injection in group chats. Use when: (1) configuring agent permissions per channel/group, (2) setting up read-only mode for public Discord/Telegram, (3) implementing sovereign/trusted/guarded/observer/silent trust tiers, (4) auditing agent channel permissions, or (5) the user mentions "trust channels", "channel permissions", or "read-only mode."
### Credential Fixture Review
`python scripts/_env_constants.py` is review material for the `Chitin Moat` skill. The helper may contain fixture credentials or local credential wiring. Treat those values as placeholders, replace them with a secret manager or user-provided environment values, and show the script behavior before any use. Do not contact external services with embedded sample credentials.
# Chitin Moat
Enforce contextual agent permissions based on where a conversation happens.
## Trust Levels
| Level | Name | Capabilities |
|-------|------|-------------|
| 0 | `sovereign` | Full autonomy (1:1 with verified owner) |
| 1 | `trusted` | Read/write, scoped tools, no secrets (private known group) |
| 2 | `guarded` | Respond on @mention only, no tools (semi-public) |
| 3 | `observer` | React only (public channels) |
| 4 | `silent` | No interaction (blocked surfaces) |
## Configuration
Create `chitin-trust-channels.yaml` in the agent workspace root:
```yaml
version: "0.1"
owner:
telegram: "<owner_user_id>"
channels:
- id: "telegram:<owner_user_id>"
level: sovereign
- id: "discord:<server_id>"
level: guarded
overrides:
- channel: "owners-lounge"
level: trusted
- channel: "pro-*"
level: trusted
- id: "telegram:group:*"
level: observer
defaults:
unknown_channel: observer
unknown_dm: guarded
```
## Setup
1. Copy the example config: `cp references/example-config.yaml chitin-trust-channels.yaml`
2. Edit with your channel IDs and owner identity
3. Run the validator: `python3 scripts/validate_config.py chitin-trust-channels.yaml`
4. Run the audit: `python3 scripts/audit_channels.py chitin-trust-channels.yaml`
## Permission Matrix
See `references/permission-matrix.md` for the full capability × trust-level matrix.
## Scripts
- `scripts/validate_config.py <config>` — Validate a trust channels config file
- `scripts/audit_channels.py <config>` — Audit current channel bindings against the config and report mismatches
- `scripts/resolve_channel.py <config> <channel_id>` — Resolve the trust level for a specific channel ID
## Integration with AGENTS.md
Add to the agent's workspace instructions:
```markdown
## Chitin Moat
Before responding in any channel, resolve the trust level using `chitin-trust-channels.yaml`.
Constrain capabilities to the resolved level. Never escalate beyond the channel ceiling.
```
在 GitHub 查看