Use cases/Contain an agent that went wrong

Contain an agent that went wrong

ProposedNo plan row yetPromptComputeRepoOpsCommitOpsContextFile

Not in the plan. Needs ID agent tokens accepted everywhere (LDG-2264, Planned).

Use case

Contain an agent that went wrong: a workflow across PromptCompute, RepoOps, CommitOps, ContextFile that starts from a person, an alert, or a PromptCompute deny pattern.

Who would use it

Security and platform leads when an agent key leaks or an agent starts doing harm.

What value it provides

One run revokes the agent version everywhere, lists what it touched, opens an incident and files a lesson.

Why it matters to the user

Agents now hold more credentials than people. Containment needs to take minutes, not an afternoon of tabs.

The problem

When an agent misbehaves, teams cannot see what it touched, and they rotate keys in each tool by hand while it keeps running.

Expected outcomes

Minutes from trigger to revoked in every product
Products that confirmed revocation: target all
Actions found in the blast radius

Each run records its outcome as met, not met or not measured.

How it runs

1. Revoke the agent version and its grantsPromptComputeBuilt

2. List what it touched (get_agent_blast_radius)RepoOpsBuilt

3. Revoke its CommitOps credentialCommitOpsBuilt

4. Revoke its tokens in every other productPromptComputePlanned

5. Open an incident and a lessonRepoOpsBuilt

Products and connections

Product	Role in this workflow	Docs
PromptCompute	Revoke the agent version and its grants; Revoke its tokens in every other product	Reference · PromptCompute docs
RepoOps	List what it touched (get_agent_blast_radius); Open an incident and a lesson	Reference · RepoOps docs
CommitOps	Revoke its CommitOps credential	Reference · CommitOps docs
ContextFile		Reference · ContextFile docs

Outside the family: GitHub, Your cloud provider (optional).

What a person approves

Action	Default risk
`access.revoke`	critical
`repoops.request_key_rotation`	high
`commitops.credential.revoke`	high

Action keys are from the PromptCompute catalog.

How to access it from an agent

An agent will run this workflow through the PromptReports MCP server or API, with an ID agent token. The runner will then call each product for it. Proposed

// MCP: https://mcp.promptreports.ai/mcp (Proposed)
start_run({
  workflow: "contain-an-agent",
  inputs: { "agent": "agt_...", "version": "agv_... or all versions" },
  dry_run: false
})
// returns { run_id, status: "running", approvals_url }
get_run({ run_id })   // poll, or subscribe to run events

curl -X POST https://api.promptreports.ai/v1/runs \
  -H "Authorization: Bearer $PROMPTREPORTS_AGENT_TOKEN" \
  -d '{"workflow":"contain-an-agent"}'

Or call the products yourself, step by step:

Product	Endpoint	Tools
PromptCompute	`https://mcp.promptcompute.ai/mcp`	`request_approval` `get_approval` `cancel_approval` `answer_approver`
RepoOps	`https://mcp.repoops.ai/mcp`	`search_brain` `get_pre_edit_context` `propose_brain_entry` `get_incident_attribution`
CommitOps	`https://commitops.ai/api/mcp`	`whoami` `list_assertions` `get_assertion` `promised_not_started`
ContextFile	Built	to confirm

Set it up

Sign in at app.promptreports.ai with PromptReports ID, and pick your org.
Connect PromptCompute, RepoOps, CommitOps, ContextFile on the Connections page. Link GitHub, Your cloud provider (optional) through the product that uses it.
Register the agent that will run it on the Agents page, and give it a version manifest.
Set the approval policy for access.revoke, repoops.request_key_rotation, commitops.credential.revoke. Start with Require approval; critical actions are never auto.
Fill in the inputs (Agent, Version) and choose the trigger: a person, an alert, or a PromptCompute deny pattern.
Start a dry run. It will count no workflow request, and it shows which steps would record missing evidence.
Start the run. Approve each gate in the approvals inbox, then read the signed record when it ends.