[Deep Agents][Skill Install] Successful install omits SHA-256 content digest from CLI output

[mercl-lau]

Description

Installing a fresh Deep Agents skill succeeds and internally returns a content digest, but the CLI output does not display the SHA-256 digest. Users receive only the validated file count and success messages, so the advertised content attestation cannot be recorded or independently compared.

Platform scope: Reproduced on Ubuntu 24.04 under WSL2; other platforms were not tested.
Regression: Unknown; earlier releases were not tested.

Environment

• OS: Ubuntu 24.04.4 LTS under WSL2
• Architecture: x86_64
• NemoClaw: v0.0.103
• OpenShell CLI: 0.0.85
• Node.js: v22.23.2
• npm: 10.9.8
• Docker: 29.5.3
• Agent: LangChain Deep Agents Code
• Inference: local Ollama

Steps to Reproduce

1. Onboard a Ready LangChain Deep Agents Code sandbox.
2. Create a valid skill directory containing SKILL.md and one additional regular file.
3. Run nemoclaw <sandbox> skill install <skill-directory>.
4. Inspect the complete stdout and stderr output.

Expected Result

The command exits 0 and reports both the uploaded file count and a SHA-256 content digest for the installed skill.

Actual Result

The command exits 0 and reports the file count and successful installation, but no SHA-256 or digest value is printed:

✓ Validated SKILL.md (name: safe-skill, 2 files)
✓ Installed 2 file(s) into the agent skill directory
✓ Skill 'safe-skill' installed
Start a new Deep Agents session to load the skill.

Source Confirmation

In v0.0.103, src/lib/actions/sandbox/skill-install.ts requires fresh.contentDigest for the success path, proving that the digest is available. The success output prints fresh.uploaded and the skill name but never prints fresh.contentDigest.

Impact

The install succeeds, but the user-visible content-attestation contract is incomplete. Users and automation cannot capture the digest as evidence or compare it across installations.