jade-recipe-8-11-tls-anon-cipher
Resolves hard-coded anonymous TLS cipher suites (DH_anon / ECDH_anon), which JDK 11 disables by default through jdk.tls.disabledAlgorithms. Agent-mode recipe: the shard contract names the files; this
SKILL.md.claude/skills/java-migration-skill-registry/8-to-11/jade-recipe-8-11-tls-anon-cipher/SKILL.md
GitHubjade-recipe-8-11-tls-anon-cipher — anonymous suites no longer negotiate
The JDK 11 java.security default for jdk.tls.disabledAlgorithms contains
anon, NULL. The JSSE reference guide states these algorithms will not be
negotiated for TLS/DTLS connections even if they are enabled explicitly by an
application. SSLServerSocket.setEnabledCipherSuites(...) with an anonymous
suite therefore no longer produces an unauthenticated handshake on JDK 11; it
produces a handshake failure at connect time. This is a runtime behavior change,
not a compile error — the code still compiles, which is why it needs a rule.
Scope of one task
You receive a shard contract. This rule is blast_class: body-local
(parallel_safe: true). Edit only that shard's editable_files;
read_only_context is read-only.
The two sites
| file | line | construct |
|---|---|---|
src/jade/src/jade/imtp/leap/SSLHelper.java |
29 | TLS_ECDH_anon_WITH_AES_128_CBC_SHA in the supportedKeys list |
src/jade/src/jade/imtp/leap/JICP/JICPSPeer.java |
224 | SSL_DH_anon_WITH_RC4_128_MD5 passed to setEnabledCipherSuites |
Both are String literals inside a String[] initializer. The declared types
(List, String[]) and every surrounding signature stay identical — that is
what makes this body-local.
What the agent may decide, and what it may not
The agent may replace an anonymous suite with another suite only when the
replacement needs no new key material and no new configuration: same
String[] shape, same call, nothing else in the file changes.
The agent must not silently move JADE from unauthenticated to authenticated TLS. Every non-anonymous suite requires a server certificate, which means a keystore, which means new configuration, a new failure mode for every existing deployment, and a decision about what JADE ships by default. That is an architectural change for the user to make, not a mechanical fix. Never invent a keystore path, never generate a self-signed certificate, never add a property.
If the only correct replacement requires authentication, report NEEDS_REVIEW
with the two options stated plainly:
- Adopt authenticated suites, and specify where key material comes from.
- Keep anonymity as an explicit, documented deployment choice by removing
anonfromjdk.tls.disabledAlgorithmsin the deployment'sjava.security, accepting that this is a downgrade a reviewer must sign off on and that it is a JVM-level setting, not a JADE source change.
Do not choose between those on the user's behalf.
Invariants
- No public signature changes — this rule is
body-local. - No new imports, no new fields, no new configuration keys.
- No change under
consumer-playground/. The consumers carry their own occurrences (JICPSConnection.java,SSLEngineHelper.java,HTTPSClientConnection.java); they are outside the workspace and outside this shard. If a JADE-side change would force a consumer edit, that isNEEDS_REVIEW, because consumers must keep passing untouched.
Verification
verify_shard.pyapplies normally here: the files survive and must compile.- A JICP-S handshake between two JADE containers on JDK 11 must complete
without
SSLHandshakeException. - All four consumers must still PASS, with no edit on their side.
Status to report
| status | when |
|---|---|
FIXED |
the anonymous suite is gone and the replacement needs no new key material or configuration |
NEEDS_REVIEW |
the only correct fix requires authentication, key material, or a JVM security-property change |
SKIPPED |
the flagged literal is not actually reachable configuration (prove it, do not assume it) |
FAILED |
the file does not compile after the edit, or a consumer stops passing |