Files
val-blog/org/capability/CAPABILITY_PROOF.md
T

120 lines
3.2 KiB
Markdown

# CAPABILITY_PROOF.md
Generated by subagent **⚙️ Anvil (Works)** on **2026-03-08 14:37 CST**.
## Objective
Produce hands-on proof of practical capability with evidence for:
1. Browser-based information retrieval via `browser` tool
2. Local command execution via `exec`
3. File artifact creation in workspace
---
## Mini-test 1 — Browser-based info retrieval (`browser` tool)
### Reproducible steps
1. Open docs site:
- Tool action: `browser.open`
- URL: `https://docs.openclaw.ai`
2. Capture page structure/content:
- Tool action: `browser.snapshot`
- `refs: aria`
### Actions performed
- `browser.open``https://docs.openclaw.ai/`
- Returned `targetId`: `F50E03EB78E5B79DC7A7A8B857A6539E`
- `browser.snapshot` on same `targetId`
### Key points captured from docs.openclaw.ai home page
- OpenClaw is presented as a **self-hosted gateway** connecting chat apps (WhatsApp, Telegram, Discord, iMessage, etc.) to AI agents.
- Positioning highlights:
- **Self-hosted** (your hardware, your rules)
- **Multi-channel** (single gateway process for multiple channels)
- **Agent-native** (tool use, sessions, memory, multi-agent routing)
- **Open source** (MIT)
- Quick start commands shown:
- `npm install -g openclaw@latest`
- `openclaw onboard --install-daemon`
- `openclaw channels login`
- `openclaw gateway --port 18789`
- Control UI default local URL shown:
- `http://127.0.0.1:18789/`
- Config path shown:
- `~/.openclaw/openclaw.json`
### Evidence excerpt
Snapshot contained visible sections/labels including:
- “What is OpenClaw?”
- “Key capabilities”
- “Quick start”
- “Dashboard”
- “Configuration (optional)”
---
## Mini-test 2 — Local tool usage (`exec`)
### Reproducible command
Executed in workspace `/Users/guchen/.openclaw/workspace`:
```bash
pwd; echo '---'; ls -1; echo '---'; node -v; echo '---'; uname -a
```
### Output (captured)
```text
/Users/guchen/.openclaw/workspace
---
AGENTS.md
BOOTSTRAP.md
HEARTBEAT.md
IDENTITY.md
MEMORY.md
SOUL.md
TOOLS.md
USER.md
memory
org
---
v22.22.1
---
Darwin MBKPro.lan 23.6.0 Darwin Kernel Version 23.6.0: Thu Mar 6 21:58:03 PST 2025; root:xnu-10063.141.1.704.6~1/RELEASE_X86_64 x86_64
```
### What this proves
- Command execution works
- Workspace filesystem is accessible
- Runtime environment details can be gathered safely
---
## Mini-test 3 — File artifact creation in workspace
### Reproducible commands
```bash
mkdir -p org/capability/artifacts
printf 'Anvil artifact created at: %s\nWorkspace: %s\n' "$(date '+%Y-%m-%d %H:%M:%S %Z')" "$(pwd)" > org/capability/artifacts/anvil_artifact.txt
shasum -a 256 org/capability/artifacts/anvil_artifact.txt
```
### Output (captured)
```text
c17d1f72cf84128a385cf5c145b09e5d229698a9f4a52632817ba2b63d93d945 org/capability/artifacts/anvil_artifact.txt
```
### Artifact content
File: `org/capability/artifacts/anvil_artifact.txt`
```text
Anvil artifact created at: 2026-03-08 14:37:49 CST
Workspace: /Users/guchen/.openclaw/workspace
```
### What this proves
- Workspace write operations succeed
- Artifact integrity can be verified with SHA-256 hash
---
## Result
All three required mini-tests were completed with command/action logs, captured outputs, and reproducible steps.