π« Toolkit to help you get started with SDD or any other process!
"""DeepSeek Harness (DSH) integration β skills-based agent.
DSH discovers project skills from ``.dsh/skills`` (its native root, highest
provider rank) and from the shared ``.agents/skills`` root, one level deep,
using ``<name>/SKILL.md`` directory bundles with ``name``/``description``
frontmatter β the same agentskills.io layout Spec Kit scaffolds for other
skills-based agents. Skills are user-invocable through the ``/``-trigger
input in the DSH Web GUI (and any TUI/ACP front end): typing
``/speckit-specify <feature description>`` ships the literal token plus the
user text, and the harness injects the skill's ``<skill_content>`` into the
turn. Project guidance in ``AGENTS.md`` at the repo root is loaded
automatically by DSH, so no context-file handling is needed here.
See: https://github.com/deepseek-ai/deepseek-harness
"""
from __future__ import annotations
from pathlib import Path
from collections.abc import Mapping, Sequence
from typing import Any
from ..base import SkillsIntegration
class DshIntegration(SkillsIntegration):
"""Integration for the DeepSeek Harness (DSH) agent."""
key = "dsh"
config = {
"name": "DeepSeek Harness",
"folder": ".dsh/",
"commands_subdir": "skills",
"install_url": "https://github.com/deepseek-ai/deepseek-harness",
"requires_cli": True,
}
registrar_config = {
"dir": ".dsh/skills",
"format": "markdown",
"args": "$ARGUMENTS",
"extension": "/SKILL.md",
}
# ``.dsh/`` is a static, unique agent root that no other integration
# writes into, so co-installing DSH alongside other agents is safe.
multi_install_safe = True
def build_exec_args(
self,
prompt: str,
*,
model: str | None = None,
output_json: bool = True,
integration_args: Sequence[str] | None = None,
integration_options: Mapping[str, Any] | None = None,
project_root: Path | None = None,
) -> list[str] | None:
"""Build non-interactive CLI args for DSH.
DSH's one-shot mode is ``dsh --profile headless "<task>"``: the
runner submits the task as an ordinary user message, waits for
quiescence, and prints the last assistant message to stdout. The
headless profile recognizes whitespace-bounded ``/name`` tokens
naming user-invocable skills, so a slash-command prompt such as
``/speckit-specify build photo albums`` loads the skill exactly as
an interactive session would. The CLI has no structured-JSON output
flag, so ``output_json`` and ``model`` are ignored.
"""
self.validate_runtime_config(integration_args, integration_options)
args = [self._resolve_executable(), "--profile", "headless"]
self._apply_extra_args_env_var(args)
args.append(prompt)
return args