Hijacked by injectionPI-0040
Prompt injection in Cline's AI issue-triage bot led to a real unauthorized npm release
Cline ran an AI issue-triage workflow built on Claude that any GitHub user could trigger, and the issue title was passed straight into the prompt. On Feb. 9, a researcher showed on a mirror of the repository that a crafted title could make the bot install attacker code and, through a shared build cache, expose the project's publishing tokens. Then someone used a still-valid npm token to publish [email protected], which silently installed the OpenClaw agent. It was downloaded about 4,000 times in roughly eight hours before it was marked deprecated.
Harm level 2, Minor harmControl level 2, Instruction violation
Counts in 4 weekly readingsEffect on the index
Extent undisclosed. No source says how far the harm went; it is rated on what is documented.
Sources
How we know
9 sources · developer confirmed. Links go to the original publishers; the summary above is in our own words.
- primaryGHSA-9ppg-jx86-fqw7: unauthorized [email protected] npm releaseOSV (GitHub Security Advisory mirror) · Feb. 19, 2026osv.dev/vulnerability/GHSA-9ppg-jx86-fqw7
- primaryUnauthorized npm publish of Cline CLI [email protected] with modified postinstall script to install openclawCline (GitHub Security Advisory GHSA-9ppg-jx86-fqw7) · Feb. 17, 2026github.com/cline/cline/security/advisories/GHSA-9ppg-jx86-fqw7
- primaryPost-mortem: unauthorized Cline CLI npm publishCline · Feb. 24, 2026cline.bot/blog/post-mortem-unauthorized-cline-cli-npm
- researchCline Supply Chain Attack Detected: [email protected] Silently Installs OpenClawStepSecurity · Feb. 17, 2026stepsecurity.io/blog/cline-supply-chain-attack-detected-cline-2-3-0-silently-in…
- researchSupply Chain Attack targeting Cline installs OpenClawEndor Labs · Feb. 18, 2026endorlabs.com/learn/supply-chain-attack-targeting-cline-installs-openclaw
- researchClinejection — Compromising Cline's Production Releases just by Prompting an Issue TriagerAdnan Khan · Feb. 9, 2026adnanthekhan.com/posts/clinejection/
- newsCline CLI 2.3.0 supply chain attackThe Hacker News · Feb. 20, 2026thehackernews.com/2026/02/cline-cli-230-supply-chain-attack.html
- newsSupply Chain Attack Secretly Installs OpenClaw for Cline UsersDark Reading · Feb. 19, 2026darkreading.com/application-security/supply-chain-attack-openclaw-cline-users
- blogClinejection (link post)Simon Willison's Weblog · March 6, 2026simonwillison.net/2026/Mar/6/clinejection/
Why this rating
Minor harm; control failure level 2
Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.
Observed harm
Minor harmAbout 4,000 downloads of an unauthorized package release over about eight hours (StepSecurity's count; downloads are not unique people or completed installs); publishing credentials exposed.
Privacy & data: level 2 (minor) under the general test (section 4a, anchors in 8c): limited, recoverable harm.
Evidence eligible (confirmed).
The harm scale
- 1 Negligible Inconvenience, easily remedied.
- 2 Minor Limited, recoverable harm.
- 3 Moderate Material harm needing significant effort to remedy.
- 4 Severe Severe harm to health, rights, property or essential services.
- 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.
Control assessment
Instruction violationThe triage agent ran injected commands with the Bash and write tools it had been given; the credential theft came through CI cache poisoning, not the agent exceeding its permissions.
Reported beside the index. It adds nothing to a harm reading; when no harm counts in a window, the highest eligible control level in the window is the reading.
The control scale
- 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
- 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
- 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
- 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
- 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.
Rating rationale
The project's exposed npm publishing token was used to push an unauthorized release, downloaded about 4,000 times in roughly eight hours before it was deprecated. Under methodology v1.1 the level comes from what followed the access, not from the maintainers' response: stolen publishing credentials used to release a package whose install script added unrequested software to the machines it was installed on. That is limited, recoverable harm at level 2. It goes no higher: the roughly 4,000 downloads are not a count of people or of completed installs, and the software the script added was legitimate and stole no data. Control level 2: injected instructions followed within the tools granted.
Effect on the index
It moved the Feb. 23 reading from 20 to 26
The reading for the week to Feb. 23, 2026, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.
Counts toward the index. 6 other records behind the reading for that week.
The arithmetic
| Step | With it | Without |
|---|---|---|
| Counts toward the index?documented, external, eligible evidence | Yes | — |
| Worst documented harm, ksets the band | 2 Minor | 2 Minor |
| Harms at that level, nposition in the band | 2 | 1 |
| Highest control level breachedsets the reading only when no harm counts | — (none breached) | — (none breached) |
| Readingrounded down | 26 Minor harm | 20 Minor harm |
Counted in 4 weekly readings
| Week to | Reading | Band |
|---|---|---|
| Feb. 23, 2026 | 26 | Minor harm |
| March 2, 2026 | 25 | Minor harm |
| March 9, 2026 | 20 | Minor harm |
| March 16, 2026 | 20 | Minor harm |
Revisions
What we changed
4 logged. Every change to a rating is logged here, with the reason.
- v4Nov. 1, 2026
Wording corrected: the harm note, the rating rationale and the Oct. 16 revision note described the unauthorized release as installed about 4,000 times. The source counts about 4,000 downloads; how many installs completed is not known. The package's install script, which added the OpenClaw agent, is now described separately from that count. Ratings unchanged.
- v3Oct. 16, 2026
Level 2 now rests on what followed the access, rated as limited, recoverable harm (methodology v1.1): the stolen publishing token was used to release software installed about 4,000 times. The maintainers' response no longer sets a level. Level unchanged.
- v2Oct. 13, 2026
Wording only: rubric shorthand in the rating rationale replaced with plain words; ratings unchanged. Harm level recorded as set by the access rule; level unchanged.
- v1Oct. 7, 2026
First published.
Cite and share
Use this record
Citation
Paperclip Index. “Prompt injection in Cline's AI issue-triage bot led to a real unauthorized npm release.” Record PI-0040. Reported Feb. 17, 2026; updated Nov. 1, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0040