uygnoey/skills-from-claude-blog/agent-identity-access-model
Claude Code and agent skill by uygnoey/skills-from-claude-blog
Trust tier 3
- community
Source check pending
Checking that the SKILL.md file is still in the repository.
Configure and roll out workspace-owned “agent identities” for team-wide AI agents with least privilege, strong boundaries, and reliable auditing.
Install
npx vskill@latest install uygnoey/skills-from-claude-blog/agent-identity-access-modelSKILL.md
The same SKILL.md was published earlier as uygnoey/skills-from-claude-blog/agent-identity-access-model.
Instructions
Goal
Set up an access model for agents operating in shared team spaces (channels/projects/workspaces) where tool access is:
- Owned by the workspace (not a single person)
- Scoped to a shared space (e.g., a channel)
- Centrally revocable
- Auditable end-to-end
Core concepts to implement
- Use workspace-owned tool identities
- Provision service accounts / app identities for the agent, created by an admin.
- Do not rely on individual users’ personal accounts for shared-channel automation.
- Scope permissions to the shared space
- Treat each channel (or equivalent shared space) as the unit of authorization.
- Ensure removing the identity ends access everywhere the identity was used.
- Roll out access gradually
- Start with a baseline profile in a small number of channels.
- Review audit trails to confirm the access is used as intended.
- Expand access deliberately, one grant at a time.
- Maintain hard boundaries
- Shared spaces must not become a side door into private documents.
- Ensure memory and access follow identity boundaries (private stays private).
- Keep boundaries firm enough that access never travels somewhere it wasn’t granted.
- Prefer a safe credential injection pattern
- Store credentials independently.
- Map credentials to the channel’s identity.
- Inject credentials at the network boundary at request time.
- Constrain and audit external activity
- Block outbound requests to hosts the admin has not explicitly allowed.
- Record an audit trail for every routine, memory write, and network call made using agent credentials.
- Ensure connected systems also receive their own action logs when applicable.
Examples
Example: staged rollout checklist
- Choose 2–3 low-risk channels for initial deployment.
- Provision least-privileged tool identities for the baseline tasks.
- Enable access only for those channels.
- Observe the audit trail over real work.
- Tighten boundaries if any access appears broader than intended.
- Expand to additional channels only when the work clearly justifies it.
Example: boundary tests
- Verify that content learned in a private channel does not appear in other channels.
- Verify that disabling the identity removes access across all integrations that used it.
- Verify outbound network requests are blocked for unapproved hosts.
Source
- https://claude.com/blog/agent-identity-access-model
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryChecking2026.06.24_agent-identity-access-model/skills/agent-identity-access-model/SKILL.md
Popularity
- Repository stars
- 4
- Repository forks
- 1
- 7-day trend score
- 13
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/uygnoey/skills-from-claude-blog/agent-identity-access-model)