← 返回 Skills 市场

Bland

Name: Bland
Author: 0xrichyrich

作者 0xrichyrich · GitHub ↗ · v1.0.0

cross-platform ⚠ suspicious

334

总下载

当前安装

版本数

在 OpenClaw 中安装

/install bland

功能描述

Place and manage AI-driven phone calls, access transcripts, recordings, call status, and configure inbound call agents via the Bland AI API.

使用说明 (SKILL.md)

Bland AI — Voice Calling Skill

Make and manage AI-powered phone calls via the Bland AI API.

Quick Start

# Make a call
bland call +14155551234 --task "Call and ask about their hours"

# Check call status
bland call-status \x3Ccall_id>

# Get transcript after call
bland transcript \x3Ccall_id>

Commands

Command	Description
`bland call \x3Cphone> [opts]`	Place an outbound AI call
`bland call-status \x3Cid>`	Get status/details of a call
`bland calls [--limit N]`	List recent calls
`bland stop \x3Cid>`	Stop an active call
`bland stop-all`	Stop all active calls
`bland recording \x3Cid>`	Get recording URL for a call
`bland transcript \x3Cid>`	Get formatted transcript
`bland voices`	List available voices
`bland numbers`	List owned inbound numbers
`bland buy-number [--area-code 415]`	Purchase an inbound number
`bland setup-inbound \x3Cphone> --task "prompt"`	Configure inbound call agent
`bland balance`	Check account balance
`bland analyze \x3Cid> --goal "question"`	AI analysis of a call

Call Options

--task "prompt"           AI agent instructions (required for useful calls)
--voice "josh"            Voice to use (default: josh)
--first-sentence "Hi!"    First thing the AI says
--from "+1234567890"      Caller ID (must own the number)
--wait-for-greeting       Wait for the other party to speak first
--wait                    Poll until call completes, then show transcript
--model "base"            Model to use (default: base)

Examples

# Restaurant reservation
bland call +14155551234 --task "Make a reservation for 2 at 7pm tonight under Joshua"

# Call and wait for result
bland call +14155551234 --task "Ask about store hours" --wait

# Screen inbound calls
bland setup-inbound +14155551234 --task "You are a call screener. Ask who is calling and why."

# Analyze a completed call
bland analyze abc123 --goal "Did they confirm the appointment?"

Environment

API Key: BLAND_API_KEY in /root/clawd/.env
API Base: https://api.bland.ai/v1
Script: /root/clawd/skills/bland/scripts/bland.sh

Notes

Phone numbers must be E.164 format: +14155551234
Calls cost money — check bland balance before heavy usage
Use --wait flag to block until a call finishes and auto-show transcript
Recording URLs are temporary — download if you need to keep them

安全使用建议

This skill's behavior (making phone calls via Bland's API) is consistent with its description, but before installing: 1) confirm you want to provide a BLAND_API_KEY and that the key is only for the Bland service; 2) inspect the script yourself (it's included) and ensure it only calls the documented API endpoints; 3) be aware it reads /root/clawd/.env (which may contain other secrets) — consider storing the Bland key in a dedicated, restricted location or setting BLAND_API_KEY in the environment instead; 4) make sure required utilities (curl, jq) are installed and that you accept the cost and privacy implications of placing outbound/inbound calls and storing transcripts/recordings; and 5) ask the publisher to update the registry metadata to declare BLAND_API_KEY and required binaries (curl, jq) to remove the mismatch. If you cannot verify these, treat the skill as untrusted.

功能分析

Type: OpenClaw Skill Name: bland Version: 1.0.0 The skill bundle provides a CLI client for the Bland AI voice calling API. It correctly loads the `BLAND_API_KEY` from `/root/clawd/.env` as documented in `SKILL.md` and uses `curl` and `jq` to interact with `https://api.bland.ai/v1`. All network communication is directed to this legitimate API endpoint. Input arguments are safely handled by `jq` to prevent shell injection. While the `--task` argument could potentially be used for prompt injection against the downstream Bland AI service, this is a vulnerability of the target AI service or the overall agent's input sanitization, not an intentional malicious act by this skill bundle itself. No evidence of data exfiltration to unauthorized endpoints, backdoors, or other malicious intent was found.

能力评估

ℹ Purpose & Capability

Name/description match the included script and SKILL.md: the script calls https://api.bland.ai/v1 to place and manage calls. However the skill metadata declares no required environment variables or binaries while both the SKILL.md and the script clearly expect a BLAND_API_KEY and rely on command-line tools (curl, jq) that are not listed as required. This mismatch is likely an oversight but is disproportionate to the stated metadata.

✓ Instruction Scope

SKILL.md and the script confine actions to calling Bland's API endpoints and local .env lookup. The runtime instructions and script do not include broad file collection, unrelated network destinations, or commands that modify other skills or system configuration beyond reading a .env file and making HTTP requests.

✓ Install Mechanism

No install spec (instruction-only with a bundled script). There are no remote downloads or archive extraction; the only code is the provided shell script, so install risk is low. However the script will be executed on the host, so missing dependency declarations matter.

⚠ Credentials

The script requires a BLAND_API_KEY (and SKILL.md documents BLAND_API_KEY in /root/clawd/.env) but the skill registry metadata lists no required environment variables or primary credential. The script also attempts to read /root/clawd/.env and a .env relative to the script; reading a root-level .env is sensitive because it may contain other secrets. Additionally, the script assumes presence of system utilities (curl, jq) that are not declared. These gaps increase the chance of accidental credential exposure or execution failures.

✓ Persistence & Privilege

The skill is not always-enabled and does not request special persistent privileges. It does not modify other skills' configs or system-wide settings in the provided code.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install bland
安装完成后，直接呼叫该 Skill 的名称或使用 /bland 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release of Bland AI — Voice Calling Skill. - Place and manage AI-powered calls via the Bland AI API. - Full CLI support for calling, transcripts, analysis, recordings, voice selection, and number management. - Includes inbound call handling, account balance checks, and automatic transcript retrieval. - Clear documentation for usage, options, and environment setup.

元数据

Slug bland

版本 1.0.0

许可证 —

累计安装 1

当前安装数 1

历史版本数 1

常见问题

Bland 是什么？

Place and manage AI-driven phone calls, access transcripts, recordings, call status, and configure inbound call agents via the Bland AI API. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 334 次。

如何安装 Bland？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install bland」即可一键安装，无需额外配置。

Bland 是免费的吗？

是的，Bland 完全免费（开源免费），可自由下载、安装和使用。

Bland 支持哪些平台？

Bland 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了 Bland？

由 0xrichyrich（@0xrichyrich）开发并维护，当前版本 v1.0.0。