A developer tool or IDE extension was compromised: what should your tabletop rehearse, and who should be in the room?
Quick answer, Nov 16, 2026As of 16 November 2026, CISA's Tabletop Exercise Packages page names no compromised-developer-tool scenario, so write it yourself. Rehearse six decisions: who had the bad version, whether it was approved, what it could reach, what evidence exists, who rotates credentials, and when the tool returns. Run senior and operational groups, including the toolchain owner, separately first, then together.
Why this scenario needs its own exercise
Most incident plans start with an attacker outside: a phishing email, an exposed server, a stolen password. This scenario starts with software your developers chose to install. An editor extension, a plugin for an integrated development environment (IDE) or an AI coding assistant ships an update, and the update is malicious. Nobody clicked anything unusual.
NIST's Secure Software Development Framework, SP 800-218 (version 1.1, February 2022), treats the developer's own setup as something to defend. Practice PO.5, "Implement and Maintain Secure Environments for Software Development", asks that those environments be "strongly protected from internal and external threats to prevent compromises of the environments or the software being developed or maintained within them." Task PO.5.2 extends that to "development endpoints", the machines of "software designers, developers, testers, builders, etc."
CISA's Tabletop Exercise Packages page does not name this case. Under "Cybersecurity Scenarios" it lists "ransomware, insider threats, phishing, and Industrial Control System (ICS) compromise". The templates under "CTEP Documents" still help: an invitation, a slide deck, a feedback form and an After Action Report. Plan to write the scenario yourself.
How this differs from two neighboring guides
A collaborator's GitHub account pushed to your repo is a live incident: read the audit log, then rotate secrets in order. Your data leaked from a vendor's cloud rehearses your data on someone else's system. This page rehearses the reverse: someone else's code running on your developers' machines. Use the first for rotation order when inject 5 fires. Use the second to choose between a tabletop and a functional exercise.
Six injects to rehearse
| Inject | The decision the room must make | Where it is written |
|---|---|---|
| 1. A tool vendor's advisory says one version of its extension was malicious | Can you list which developers had that version installed today, without asking each of them? | CSF 2.0 ID.AM-02 (Asset Management), "Inventories of software, services, and systems managed by the organization are maintained"; SSDF PO.3.1, "Specify which tools or tool types must or should be included in each toolchain" |
| 2. Was it approved? | Was the extension on an approved list, or installed freely? Who decides now? | CSF 2.0 PR.PS-05 (Platform Security), "Installation and execution of unauthorized software are prevented"; SSDF PO.5.2 Example 2, "least functionality" and "the principle of least privilege" |
| 3. What could it reach? | From that endpoint, could the tool touch source code, signing keys, cloud credentials or production? | SSDF PO.5.1 Example 2, separating environments "from each other and from production environments"; SSDF PS.1.1, code stored "based on the principle of least privilege" |
| 4. What evidence exists? | Which logs exist on developer endpoints, and who pulls them before machines are rebuilt? | SSDF PO.5.2 Example 3, "Continuously monitor the security posture of all development endpoints"; CSF 2.0 DE.CM-09 (Continuous Monitoring) |
| 5. Credentials and the vendor | If tokens sat on affected machines, who rotates them, and who talks to the tool vendor? | CSF 2.0 RS.MA-01 (Incident Management), "The incident response plan is executed in coordination with relevant third parties once an incident is declared" |
| 6. When does it come back? | Who signs off before the tool, or its replacement, returns to developer machines? | CSF 2.0 PR.PS-02, "Software is maintained, replaced, and removed commensurate with risk"; SSDF PO.3.2 Example 7, "Regularly verify the integrity and check the provenance of each tool" |
Run inject 1 first and give it time. If the room's answer is to ask around in chat, the exercise has found its first gap. It is an inventory gap, not a response gap.
Who should be in the room
NIST SP 800-84, section 4.2.4 "Identify the Participants", sets the rule: participants "should be comprised of the personnel with roles and responsibilities identified in the plan". Its footnote 13 adds that "it may be necessary to make participation mandatory for designated personnel." Use that for the engineers who own the toolchain.
Do not put everyone in one room on the first run. Section 4.2.2, "Determine the Scope", says "senior-level teams and operational-level teams should participate in separate tabletop exercises initially". Section 4.2.4 adds that once both groups have exercised separately, "it is appropriate to conduct a combined session".
NIST SP 800-61r3 (April 2025), section 2.2 "Incident Response Roles and Responsibilities", gives example roles. Mapped to this scenario:
- Leadership, who "may have decision-making authority on high-impact response actions, such as shutting down or rebuilding critical services". Here, that means pausing releases or rebuilding developer machines.
- Incident handlers, who the section says can be on staff, on contract, or available when needed from an outside provider.
- Technology professionals, which the section says includes "software developers". Seat the engineering lead and whoever owns the toolchain under SSDF PO.3.
- Legal, who "can also review contracts with technology suppliers and other third parties when there are incident response implications". Here, that means the tool vendor's terms.
- Public affairs, because "the media learns of incidents through alternate sources".
- Asset owners, for the products whose code the affected developers touched.
The tool vendor cannot sit in your room. Have a staff member read its advisory aloud in stages as the injects advance.
If packages came in with the tool
If the compromised tool added packages to a project, an AI assistant can check the lockfile with check_dependencies on the Distributed Risk MCP server. It checks package versions against OSV.dev, including OpenSSF's malicious-package reports. It does not read editor extensions. A version with no hit is not a statement that it is safe.
Before you book a facilitator
Ask three things. Will they write the injects around the tools your developers actually run? Will someone in the room be able to read your build pipeline? Will the after action report name each gap against an SSDF practice ID, so engineering can act on it?
For a facilitator, a scenario built around your toolchain, or a response firm on retainer, brief an advisor and we arrange a partner firm.
Questions
We don't sell software. Does this apply? If you have an internal development team, yes. The SSDF "Audience" section counts "internal development teams" among software producers "regardless of size, sector, or level of maturity."
Will a tabletop stop a malicious extension? No exercise guarantees that. It shows whether your inventory, approvals and logs would let you answer the six questions above. Closing the gaps is a separate decision.
Is this legal advice? No. This page is a procedural guide drawn from the public record. Counsel decides what your disclosure and notification obligations are.