Hi maintainers — first of all, thanks for HarnessKit; managing 12 agents from one dashboard is great.
I'd like to use HarnessKit to manage PenguinHarness on my machine (it's a local agent harness that keeps multiple agent states under one data root). Right now it isn't among the supported agents, so I'm planning to add a penguin adapter, and wanted to ask before writing it whether that contribution would be welcome.
What PenguinHarness looks like on disk
<penguin-data-root>/
agents/
<agent_id>/
agent_state/
AGENTS.md
skills/<skill-name>/SKILL.md
memory/*.md
schedule/ ...
...
Key points:
- One data root, multiple "agent states" (one per agent id) — structurally similar to how the dsh adapter handles multiple profiles under one home, so that pattern can be reused.
- Each agent state has its own
AGENTS.md, skills/ (standard SKILL.md layout), memory/, etc.
- The harness itself is not a "big-name" CLI like Claude/Codex — same situation as dsh was before it got an adapter.
Proposed scope for a first contribution
- New adapter
crates/hk-core/src/adapter/penguin.rs implementing AgentAdapter:
detect() checks the data root exists (also respect an env override, like DSH_HOME);
skill_dirs() enumerates agents/*/agent_state/skills;
- config files:
AGENTS.md / memory files tracked like other agents' global config files.
- MCP / plugin / hooks declared unsupported initially (Penguin's formats differ; can be added later, mirroring how some existing adapters started minimal).
- Follow
CONTRIBUTING.md: build + tests green, update the capability matrix and AGENT_ORDER.
Questions
- Would you be open to a
penguin adapter contribution (or would you rather keep the supported list at the current 12)?
- Any requirements for a new agent adapter beyond what
CONTRIBUTING.md already asks for (docs, screenshots, verification against a real install, etc.)?
Happy to do the work — just want to confirm it's welcome before spending time on it. Thanks!
Hi maintainers — first of all, thanks for HarnessKit; managing 12 agents from one dashboard is great.
I'd like to use HarnessKit to manage PenguinHarness on my machine (it's a local agent harness that keeps multiple agent states under one data root). Right now it isn't among the supported agents, so I'm planning to add a
penguinadapter, and wanted to ask before writing it whether that contribution would be welcome.What PenguinHarness looks like on disk
Key points:
AGENTS.md,skills/(standardSKILL.mdlayout),memory/, etc.Proposed scope for a first contribution
crates/hk-core/src/adapter/penguin.rsimplementingAgentAdapter:detect()checks the data root exists (also respect an env override, likeDSH_HOME);skill_dirs()enumeratesagents/*/agent_state/skills;AGENTS.md/ memory files tracked like other agents' global config files.CONTRIBUTING.md: build + tests green, update the capability matrix andAGENT_ORDER.Questions
penguinadapter contribution (or would you rather keep the supported list at the current 12)?CONTRIBUTING.mdalready asks for (docs, screenshots, verification against a real install, etc.)?Happy to do the work — just want to confirm it's welcome before spending time on it. Thanks!