CtrlK
BlogDocsLog inGet started
Tessl Logo

ainativedev/latest-aidevcon-speakers-london-2026

AI Native DevCon 2026 London — all conference sessions as interactive skills

66

Quality

83%

Does it follow best practices?

Impact

—

No eval scenarios have been run

SecuritybySnyk

Risky

Do not use without reviewing

Overview
Quality
Evals
Security
Files

SKILL.mdtalk-selajev-docker-sandboxes-agents/

name:
talk-selajev-docker-sandboxes-agents
description:
Use when the user asks about Oleg Selajev's AI Native DevCon talk on sandboxing local AI agents: hard isolation, file-sharing boundaries, network policy, sensitive-value isolation, audit expectations, and safe team rollout. This bundle is safety-redacted and avoids setup instructions.
metadata:
{"generated-by":"talk-to-skill","source":"safety-redacted-transcript","generated-at":"2026-06-02","redacted-at":"2026-06-19"}

Docker Sandboxes For Agents - Oleg Selajev

Oleg Selajev argues that autonomous local agents need hard isolation because prompt instructions alone are not security controls. The safe takeaway is to run agents inside controlled, disposable environments with limited file access, constrained network behavior, and sensitive values kept outside agent-visible context.

Safety Scope

  • Provide conceptual guidance and policy checklists.
  • Do not provide setup commands, manifests, registry references, sensitive-value routing recipes, or step-by-step configuration.
  • If the user asks for omitted mechanics, explain that the public bundle is redacted and provide a safe architecture checklist.

Read Order

  1. Use outline.md for the talk thesis and safe concept map.
  2. Use transcript.md for the redacted summary.
  3. Use quotes.md for safe themes.

How To Help

When answering factual questions:

  1. Keep the answer short and concept-level.
  2. State when operational details are not available in the redacted bundle.
  3. Focus on enforceable isolation, not prompt-only policy.

When applying the talk to the user's setup:

  1. Identify what the agent can read.
  2. Identify what the agent can write.
  3. Identify where network traffic can go.
  4. Identify how sensitive values are kept out of model-visible context.
  5. Identify logs, review points, and recreation strategy.

Safe Outputs

  • Agent isolation checklist.
  • Team rollout checklist.
  • Risk-model summary.
  • Sandbox adoption plan.

talk-selajev-docker-sandboxes-agents

outline.md

quotes.md

SKILL.md

transcript.md

README.md

tile.json