Yichen Jiang
8c2970e70e
Review found five real defects in the configuration-source work, all confirmed against the code rather than argued: 1. The note claimed --config outranks settings.yaml. It does not: the settings seam registers a plugin's cordis entry config as the `base` layer and the user section layers over it, and the seam cannot tell a shipped value from a --config one. The note now states shipped reality and names --config-replace as the lever for a deployment that must win. Separately, a literal `apiKey` in settings outranked both the environment and .credentials.yaml — the field is removed, so configuration carries a reference and nothing else. 2. DEEPSEEK_SEARCH_BASE_URL was functionally deleted: the shipped inline went away without the provider learning to read it. It now resolves from the environment snapshot, as the README always claimed. 3. The bootstrap deny list missed the interpreter start-up hooks. BASH_ENV is the sharpest: `bash -c` sources it on every bash tool call, so a project .env could run a file of its choosing before every command. The list now covers BASH_ENV and its per-language siblings, the Git hook commands, and the remaining preload and CA variables, organised by what a variable does rather than which runtime owns it. 4. YAML parse errors quoted the offending source line — which in a credentials document is the secret — into boot stderr and the watcher's logger. Only the error code and position are reported now, in credentials-local and settings-local alike, pinned by a test that asserts the secret is absent. 5. 0600 governed only files the harness wrote. A hand-created 0644 document was read normally. POSIX now checks the mode before reading contents, at boot and on every reload; Windows has no mode to inspect and is skipped rather than faked. The project a session is launched in is trusted by default, with no prompt and no stored trust record: it may supply its own endpoint, ordinary variables, and a key ranked below the managed store. Trust stops at the harness itself — a discovered file still cannot set DSH_PERMISSION_MODE, PATH, BASH_ENV, or the rest, because those take effect with no user action, before any turn, outside the permission policy and the sandbox.
Examples
English | 中文
Runnable demos (not workspaces) that showcase how the harness is wired. Each example is a thin leaf: either a cordis.yml tree that picks swappable backends and loads one app package, or an overlay — a patch list dsh --config applies over the shipped composition (apps/cli/config/base.cordis.yml plus a surface overlay). Bundled compositions live in @deepseek-ai/dsh-cli-demo, @deepseek-ai/dsh-acp-demo, and their shared @deepseek-ai/dsh-agent-spine-demo bundle; the dsh surfaces use flat config trees instead. There is no start.ts; the terminal demo:* scripts boot through the dsh CLI, and the headless/ACP scripts invoke the cli-demo/acp-demo bins.
mcp-memory
Three default-off reference overlays connect a memory MCP server through the generic MCP client. Pick one file and pass it to dsh --config; DSH does not install or configure the upstream memory system. See mcp-memory/README.md for pinned prerequisites, identity mapping, the shared optional prompt, and the write → fresh-session recall → use verification recipe.
headless-agent
A non-interactive agent demo that accepts one positional task, runs one complete model/tool turn on the @deepseek-ai/dsh-cli-demo app, persists a fresh session, prints text, json, or stream-json, and exits.
Run with: pnpm run demo:headless "task" (needs DEEPSEEK_API_KEY). See headless-agent/README.md for the output contract, safety boundaries, and snapshot suite.
jsonrpc-agent
An unattended coding agent driven through the Python SDK: JSON-RPC stdio, foreground-only bash, read / write / edit, one foreground subagent, todo_write, JSONL persistence, and compaction. It excludes terminal UI, stdout logging, approvals, skills, and background task controls. See jsonrpc-agent/README.md.
web-cordis
The self-referential demo: the coding spine plus @deepseek-ai/dsh-tool-cordis, whose three tools (cordis_inspect / cordis_mount / cordis_unmount) let the agent inspect the current DSH process, mount model-written temporary Plugins (an event listener, a brand-new tool, or a service another temporary Plugin injects), and unmount them again. These Plugins exist only in memory and share one internal cordis-dynamic fiber subtree; ctx.fs/ctx.web ride along provider-only as capabilities they can use.
Run the browser UI at http://127.0.0.1:3081 with pnpm run demo:cordis, or the ACP server with pnpm run demo:cordis acp (both need DEEPSEEK_API_KEY). See the toolset Agent Note for the design and sandbox caveats.
acp-agent
An agent exposed as an Agent Client Protocol (ACP) automation server over JSON-RPC stdio, via @deepseek-ai/dsh-acp-demo. Programmatic clients create fresh sessions, send text prompts, consume committed assistant text, answer one-shot permission requests, and cancel work. It owns the ACP keyless snapshot suite.
Run with: pnpm run demo:acp (needs DEEPSEEK_API_KEY); pnpm run demo:code-mode boots the same server in Code Mode via the code-mode.cordis.yml overlay. See acp-agent/README.md for the protocol and snapshot-test contracts.
The default cordis.yml composes @deepseek-ai/dsh-sandbox-local, @deepseek-ai/dsh-bash-sandbox, and @deepseek-ai/dsh-user-approval. workspace-write confines bash and filesystem mutations to each session workspace; a wider retry becomes a one-shot machine permission request over ACP.