Begin by choosing the narrowest scope that fits the experiment. Put a test copy in a disposable project .dsh/skills directory if the Skill is only needed for one workspace. Confirm that DSH discovers the directory and that /dot-skill is available before adding real material. The upstream README is the evidence for these paths; AgentPlugin.net has not performed that discovery check.
Dot Skill
Dot Skill distills supplied material into reusable persona and work Skills that DeepSeek Harness can discover from its native skills directories.
Review snapshot
- Source
- Unknown
- Install
- Not tested
- Runtime
- Not tested
Static compatibility snapshot dated 2026-08-14; install and runtime remain independent checks.
Install / upstream documentation
Use the pinned upstream documentation
No version-pinned install command is recorded for this listing. Use the pinned README as the source of truth; AgentPlugin.net does not invent missing steps, Profiles, or recovery instructions.
The pinned README documents cloning into ~/.dsh/skills/dot-skill or .dsh/skills/dot-skill, but does not publish a version- or commit-pinned DSH install command. Follow that pinned source carefully rather than treating this directory listing as an installer.
Source: Pinned upstream install sectionPlugin brief
What Dot Skill adds to DeepSeek Harness
Dot Skill is a DeepSeek Harness plugin distributed as a filesystem Skill rather than a Cordis Bundle. Its pinned README says that DeepSeek Harness discovers it natively when the repository is placed in either the global ~/.dsh/skills/dot-skill directory or a project .dsh/skills/dot-skill directory. In a DSH session, the documented entry point is /dot-skill. This listing records that declared distribution path; it does not imply that the repository contains a package manifest or a DSH patch.
The project describes a way to distill source material into Skills that represent working practices, a persona, or both. It organizes generated outputs around colleague, relationship, and celebrity families. The stated goal is not simply style imitation: the repository describes separate prompt and research structures for work patterns, personality, decision making, and source analysis. That can be useful for preserving a team playbook or building a reusable research aid, but it also makes the quality and legitimacy of input material central to the result.
For DeepSeek Harness users, the practical value is that an ordinary Skill directory can be evaluated without adding a Web-profile bundle or a browser client. A project-local placement may make sense when the material and resulting Skill belong only to one repository; a global placement has broader reach and therefore deserves more caution. The README names both locations but does not prescribe a security model, access policy, or retention policy for your inputs.
The repository is MIT-licensed, yet that license covers the code and project files, not rights in chat logs, emails, documents, screenshots, or public material you supply. Treat Dot Skill as a content-processing workflow. Obtain permission where appropriate, minimize sensitive material, inspect generated files before sharing them, and avoid presenting a generated persona as a real person or as an authoritative record of what that person would say.
Evaluation workflow
Evaluate Dot Skill with clear boundaries
Separate source collection from generation. The README lists materials such as chat history, documents, email exports, screenshots, and manual input. Before any tool processes them, remove secrets, unrelated third-party conversations, credentials, and data whose use you cannot justify. Keep a source log outside the generated Skill so you can later explain provenance, correct errors, or delete the output if consent changes.
Use the generated result as a draft operating aid, not as an identity claim. Check whether its suggested work practices are supported by the source material, whether it overstates confidence, and whether its tone crosses a boundary. A convincing voice does not validate facts. For workplace use, require a human reviewer for recommendations that affect people, policy, spending, or production changes.
Review the repository tooling before invoking collectors or version-management scripts. The README describes optional integrations for Feishu, DingTalk, Slack, email parsing, research downloads, and generated-Skill publishing. Those paths may touch network services, local exports, or credentials. Keep backups and access controls under your own control, and reread the pinned README whenever upstream changes its workflow.
Verification ledger
Evidence, not installation claims
Source
UnknownThis is a Skill distribution. The pinned commit has no package.json or cordis.patch.yml to validate as a DSH Bundle; its README and SKILL.md are the available integration evidence.
Checked Aug 14, 2026
Install
Not testedNo clean DSH filesystem-discovery install was run by AgentPlugin.net.
Runtime
Not testedNo /dot-skill session smoke test was run by AgentPlugin.net.
| Layer | Result | What was checked |
|---|---|---|
| Source | Unknown | This is a Skill distribution. The pinned commit has no package.json or cordis.patch.yml to validate as a DSH Bundle; its README and SKILL.md are the available integration evidence.[Aug 14, 2026] |
| Install | Not tested | No clean DSH filesystem-discovery install was run by AgentPlugin.net. |
| Runtime | Not tested | No /dot-skill session smoke test was run by AgentPlugin.net. |
Source or manifest review can be recorded while Install and Runtime remain not-tested. Treat these as separate verification layers.
Execution surface
Prepare, build, and code-execution review
prepare
not recorded
No review record is available for this execution surface. Inspect the upstream repository before running it.
build
not recorded
No review record is available for this execution surface. Inspect the upstream repository before running it.
code execution
review recorded
The repository includes Python and shell-adjacent tooling for collection, generation, research, publishing, and version management. Review scripts before executing them.
network
review recorded
Optional source collection and research workflows may interact with Feishu, DingTalk, Slack, email, or video-related services.
configuration
notice recorded
Global Skill placement broadens availability across DSH workspaces; project placement limits scope but does not itself protect source content.
Declared scope
Capabilities and prerequisites
Capabilities
- Creates persona and work-oriented Skills from supplied source material
- Offers colleague, relationship, and celebrity-oriented generation families
- Documents native DSH filesystem Skill discovery and the /dot-skill entry point
- Includes source-collection and version-management tooling in the repository
Prerequisites
- A DeepSeek Harness installation with native filesystem Skill discovery
- A deliberate choice of global ~/.dsh/skills or project .dsh/skills scope
- Source material you are entitled to collect, process, and reuse
- Python 3.9 or later for the repository tooling, as declared by upstream
Frequently asked questions
Dot Skill FAQ
What is Dot Skill?
Dot Skill is a repository-distributed Skill that its README says DeepSeek Harness can discover from native filesystem Skill directories. It is intended to turn supplied material into reusable persona and work-oriented Skills.
Is Dot Skill a DSH Bundle?
No bundle manifest or cordis.patch.yml is evidenced at the pinned commit. This directory labels its distribution as a Skill, which is a different DSH integration path.
Where does upstream say to place it for DeepSeek Harness?
The pinned README names ~/.dsh/skills/dot-skill for global use and .dsh/skills/dot-skill for a project-local installation. Consult that source for the current cloning guidance.
What does the Skill do with source material?
Upstream describes prompts and tools for deriving work and persona-oriented Skills from materials such as messages, documents, exports, images, and manual input. Results should be reviewed as generated interpretations.
Can I use private chat logs or workplace material?
Only if you have a legitimate basis and appropriate permission. The code license does not grant rights in input material, and generated outputs can reproduce sensitive patterns or personal information.
Has AgentPlugin.net tested Dot Skill in DeepSeek Harness?
No. The listing is based on static inspection of the pinned README, SKILL.md path, license, and commit. Installation and runtime remain untested.