Responsible Disclosure
Kilimanjaro welcomes good-faith security research. If you find a vulnerability in the platform — the API surface, the Console, the SDK packages, the Kibo + Gate components, the extension framework, any published extension (Trailblazer-class or third-party), or any platform-managed infrastructure — please report it through the channels below. We commit to acknowledging reports, triaging them on a defined timeline, and crediting researchers whose work helps the platform.
1. In scope
The following surfaces are in scope for the programme:
| Surface | Scope |
|---|---|
| Platform API | api.k9o.io (root domain) plus per-org subdomains on both two-namespace classes — API class (<orgSlug>.api.k9o.io) and CDN class (<orgSlug>.cdn.k9o.io) |
| Console | console.k9o.io and the embedded Console surfaces |
| Marketing + Dev Hub | kilimanjaro.dev, docs.kilimanjaro.dev (limited scope — content-only) |
| SDK packages | @k9o/core, @k9o/kibo, @k9o/gate, @k9o/gate-studio, and per-language SDKs as they publish |
| Extension framework + all extensions | All published extensions — Trailblazer-class (Paperless, Context, Lookout, Gate Studio) and any third-party extension in the Extensions Hub — plus the extension framework itself |
| Kibo + Gate components | Hardware-attested enclave components, custody-transition flows, hardware-attestation chain |
Anything that can compromise customer data, customer keys, customer audit-log integrity, identity-token forgeability, or the platform's structural sovereignty claims is in scope.
2. Out of scope
The following are explicitly excluded:
- Third-party services — sovereign-cloud substrate providers (Hetzner, OVH), customer-managed KMS endpoints (AWS KMS, GCP Cloud KMS, Azure Key Vault, HashiCorp Vault), customer SIEM endpoints. Report to those vendors directly.
- Social-engineering attacks against Kilimanjaro employees or contractors.
- Physical attacks against Kilimanjaro infrastructure (the sovereign-cloud substrate providers handle physical security per their own programmes).
- Volumetric attacks (DDoS, rate-limit bypass at scale) — the platform's DDoS posture is operational, not a research target.
- Findings that require already-compromised customer credentials — if the attack requires the customer's API key or active session, it's not a platform vulnerability.
- Self-XSS in the Console (XSS that requires the attacker to have already authenticated as the victim).
- Spam / abuse reports — use the standard abuse-reporting channel.
- Issues in customer-owned applications running on the platform — those are the customer's responsibility; the platform's responsibility ends at the API surface contract.
3. How to report
Email [email protected] with a description of the vulnerability, reproduction steps, expected vs actual behaviour, and the impact you've assessed. Encrypted communication is supported via the platform's PGP key — fingerprint published at https://kilimanjaro.dev/.well-known/security.txt. Reports submitted in clear are accepted; for high-impact findings the encrypted channel is preferred.
Include:
- Description — what the vulnerability is, in your own words.
- Reproduction steps — minimal repro that demonstrates the vulnerability.
- Impact assessment — what an attacker could do with this; whose data is at risk; what platform property is compromised.
- Suggested remediation (optional but appreciated).
- Researcher contact — preferred name for credit (or "anonymous"), affiliation if disclosing one, contact channel for follow-up.
Please do not disclose the vulnerability publicly until we've coordinated remediation per the timeline below.
4. Response timeline
| Phase | Commitment |
|---|---|
| Acknowledgement | Within 2 business days of receipt |
| Triage outcome | Within 5 business days (severity classification, scope confirmation, initial remediation plan) |
| Remediation target | Per severity — Critical: 7 days; High: 30 days; Medium: 90 days; Low: best-effort within the next major platform cycle |
| Coordinated disclosure | Per researcher and platform mutual agreement; default 90 days post-remediation |
Severity classification follows CVSS 4.0; classification rationale shared with the researcher at triage outcome. Disagreements on classification are negotiated; the researcher's input is weighted but not dispositive (the platform's customers' threat model is the final reference).
5. Safe harbour
Researchers acting in good faith — within the scope above, without violating customer data, without service disruption, without monetary extortion — are protected from legal action by Kilimanjaro. We will not pursue civil action or criminal complaints against researchers whose conduct adheres to these terms.
"Good faith" specifically excludes:
- Accessing customer data beyond what's strictly necessary to demonstrate the vulnerability.
- Modifying or destroying customer data.
- Exfiltrating customer data off-platform.
- Pivoting from a platform finding into a customer's application or infrastructure.
- Disrupting platform availability for other customers.
If you're unsure whether a research approach crosses the safe-harbour boundary, surface the question to [email protected] before proceeding — we'd rather coordinate than litigate.
6. Recognition
For researchers who've helped the platform:
- Hall of fame. A public list of researchers credited for accepted findings, with consent. Anonymous reports honoured as "anonymous" on the list.
- Public CVE assignment. For findings warranting a CVE (per CVE-Numbering-Authority criteria), Kilimanjaro coordinates with MITRE on assignment + the researcher's preferred credit format.
- Bug-bounty rewards. For accepted findings of Medium severity and above, a bounty programme operates per a published reward schedule. Bounty eligibility depends on severity, novelty, and the researcher's adherence to the safe-harbour terms above. The full reward schedule is shared with researchers at triage outcome.
The bounty programme is operated directly by the platform, not through a third-party bounty service. Payment terms and tax-reporting requirements are surfaced at first qualifying report.
7. What we won't do
- We won't ask you to sign an NDA before disclosing a vulnerability.
- We won't retroactively change the safe-harbour terms.
- We won't penalise good-faith researchers for inadvertent mistakes (mis-classifying scope, accidentally exceeding the necessary access; surface the mistake and we coordinate the remediation).
- We won't ignore reports. Every submission gets an acknowledgement within 2 business days; if the timeline slips, surface it to [email protected]/escalation.
8. See also
- Security → Compliance — compliance posture that the disclosure programme composes with.
- Security → Threat model — the four-layer anti-injection posture and trust-boundary model.
- API → Security — audit log + RBAC + transport security mechanics.
- Guides → Sovereignty — structural sovereignty model.
- Guides → Observability — audit-log evidence surfaces.