Load model-specific repo instructions (CLAUDE.md for Claude only, GEMINI.md for Gemini only)

[enyst]

Summary

• Implement model-specific repository instruction loading. CLAUDE.md is included only when using Anthropic/Claude models; GEMINI.md is included only when using Gemini models.
• Wire LLM model identifiers into AgentContext rendering so the filter can be applied without breaking backward compatibility.
• Add focused unit tests validating the behavior for Claude, Gemini, other providers, and canonical-name routing.

Motivation
This change addresses issue #1282 by ensuring provider-specific instruction files are applied only for the relevant model family. This avoids cross-provider bleed of rules and aligns behavior with prior discussions and PR #1280.

What changed

• openhands-sdk/openhands/sdk/context/agent_context.py
  • get_system_message_suffix now accepts optional llm_model and llm_model_canonical parameters.
  • Applies a simple vendor check to filter out CLAUDE.md/GEMINI.md skills unless the selected model matches Anthropic(Claude)/Gemini respectively.
• openhands-sdk/openhands/sdk/agent/base.py
  • Pass LLM.model and LLM.model_canonical_name into AgentContext.get_system_message_suffix(), preserving previous behavior for other skills and custom suffixes.
• tests/sdk/context/test_agent_context_model_specific.py (new)
  • Adds unit tests to confirm only the matching provider's instructions are included.

Design notes

• Filtering is done at render time (AgentContext) to keep Skill loading simple and preserve existing patterns.
• Detection is provider-name based and considers both model and model_canonical_name for proxy scenarios.
• Backward compatibility maintained: if model is not provided, behavior is unchanged.

Testing

• Added new tests: tests/sdk/context/test_agent_context_model_specific.py
• Ran the full AgentContext tests to ensure no regressions: all pass locally.

Closes #1282

Co-authored-by: openhands [email protected]

@enyst can click here to continue refining the PR

---

Agent Server images for this PR

• GHCR package: https://github.com/OpenHands/agent-sdk/pkgs/container/agent-server

Variants & Base Images

Variant	Architectures	Base Image	Docs / Tags
java	amd64, arm64	eclipse-temurin:17-jdk	Link
python	amd64, arm64	nikolaik/python-nodejs:python3.12-nodejs22	Link
golang	amd64, arm64	golang:1.21-bookworm	Link

Pull (multi-arch manifest)

# Each variant is a multi-arch manifest supporting both amd64 and arm64
docker pull ghcr.io/openhands/agent-server:39e9569-python

Run

docker run -it --rm \
  -p 8000:8000 \
  --name agent-server-39e9569-python \
  ghcr.io/openhands/agent-server:39e9569-python

All tags pushed for this build

ghcr.io/openhands/agent-server:39e9569-golang-amd64
ghcr.io/openhands/agent-server:39e9569-golang_tag_1.21-bookworm-amd64
ghcr.io/openhands/agent-server:39e9569-golang-arm64
ghcr.io/openhands/agent-server:39e9569-golang_tag_1.21-bookworm-arm64
ghcr.io/openhands/agent-server:39e9569-java-amd64
ghcr.io/openhands/agent-server:39e9569-eclipse-temurin_tag_17-jdk-amd64
ghcr.io/openhands/agent-server:39e9569-java-arm64
ghcr.io/openhands/agent-server:39e9569-eclipse-temurin_tag_17-jdk-arm64
ghcr.io/openhands/agent-server:39e9569-python-amd64
ghcr.io/openhands/agent-server:39e9569-nikolaik_s_python-nodejs_tag_python3.12-nodejs22-amd64
ghcr.io/openhands/agent-server:39e9569-python-arm64
ghcr.io/openhands/agent-server:39e9569-nikolaik_s_python-nodejs_tag_python3.12-nodejs22-arm64
ghcr.io/openhands/agent-server:39e9569-golang
ghcr.io/openhands/agent-server:39e9569-java
ghcr.io/openhands/agent-server:39e9569-python

About Multi-Architecture Support

• Each variant tag (e.g., 39e9569-python) is a multi-arch manifest supporting both amd64 and arm64
• Docker automatically pulls the correct architecture for your platform
• Individual architecture tags (e.g., 39e9569-python-amd64) are also available if needed

[github-actions]

Coverage Report •

File	Stmts	Miss	Cover	Missing
openhands-sdk/openhands/sdk/context/skills
skill.py	250	202	19%	97–98, 100–108, 110, 113, 117–121, 126, 129, 133, 145–146, 151, 158, 171, 174–175, 177, 181, 184–186, 189–191, 193–195, 198, 201, 204–206, 212, 214–220, 223, 231–232, 240–243, 255–262, 268–269, 272–273, 275, 281–282, 284, 292–302, 309–311, 319–321, 343–344, 346–347, 350, 353–354, 357–361, 364–366, 369–370, 372–374, 382, 385, 387, 391–392, 395–396, 398–400, 402–403, 405, 409, 430–431, 433–436, 438–440, 443–447, 449, 453–454, 456, 459, 473–475, 493, 495–499, 506, 513–517, 522–524, 526, 541, 543, 545–549, 552–555, 589, 591, 593–594, 596–598, 601–604, 607, 609, 612–614, 618–624, 626–627, 629, 632
TOTAL	12427	5670	54%

[openhands-ai]

Looks like there are a few issues preventing this PR from being merged!

• GitHub Actions are failing:
  • Pre-commit checks

If you'd like me to help, just leave a comment, like

@OpenHands please fix the failing actions on PR #1328 at branch `openhands/model-specific-repo-instructions-1282`

Feel free to include any additional details that might help me get this PR into a better state.

_{^{You can manage your notification settings}}