Multi-dimensional quality scoring for DeepSeek Harness plugins that scores a repo or npm package across install success, maintenance activity, documentation completeness, security scan, and protocol compliance using real CLI evidence, and produces a JSON or Markdown leaderboard report.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-score
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:PerryLink/dsh-score
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
dsh-score
Turn any repository or npm package into a five-dimension quality scorecard with evidence links — install, maintenance, docs, security, compliance, every point traceable.
A local scorer for DeepSeek Harness: it collects evidence with dsh-v0.2.2-rc.4(2026-11-18);gh api and npm view only — it never installs, builds, or runs the target — and records no-evidence instead of inventing a number.
Install · Configuration · FAQ · 简体中文 · English · Español · Português · हिन्दी
dsh plugin --profile <name> add dsh-score
Community plugin, not affiliated with DeepSeek. Canonical repository: https://github.com/PerryLink/dsh-score (Gitee is a mirror).
What it is
dsh-v0.2.2-rc.4(2026-11-18);dsh-score is a local quality scorer for DeepSeek Harness: give it a GitHub repository or an npm package and it runs the five dimensions — install success, maintenance, documentation, security, and compliance — then returns a structured score card (schema: "dsh-score/v1") or a JSON + Markdown leaderboard. Every conclusion comes from real gh/npm output and every piece of evidence carries source, sanitized detail, and observedAt; a dimension that gathers no evidence reports no-evidence (score 0, excluded from the weighted total) instead of a guessed number. Scoring is on demand only — nothing is installed, built, or executed. This plugin is the local on-demand scorer; the public evidence layer lives in dsh-attention-index.
Three-step setup
Install (pick one channel):
# git channel (latest main; equivalent to git+https://github.com/PerryLink/dsh-score.git) dsh plugin --profile web add github:PerryLink/dsh-score#<commit-sha> # npm channel (published releases) dsh plugin --profile web add dsh-scoreThe first
addthrough the git channel fails because pnpm blocks the package'spreparebuild; copy the exact key pnpm printed into the profile'spnpm-workspace.yamland re-run:allowBuilds: 'dsh-score': truePrebuilt packages from the npm channel need no build allowance.
Minimal config (the defaults work; add keys only to change thresholds or weights):
- insert: - id: score name: dsh-score config: weights: install: 25 maintenance: 20 documentation: 20 security: 20 compliance: 15Verify it mounts:
dsh --profile web --dump-config | grep -A2 'id: score'> /score owner/repo [1/1] owner/repo → B (88/100, healthy (weighted total 88/100))Observed from
src/batch.ts:70
Uninstall: dsh plugin --profile web remove dsh-score (or remove the row from the profile patch).
dsh-scoredoes not bundle the storage stack. When the composition has nostorageDomain, the tools still work and score persistence is disabled with a logged reason; the publisheddsh-basebundle mounts storage-domain from0.1.2-rc.1on (verified against the0.1.2-rc.1and0.1.5-alpha.1tarballs), so persistence is active on the published line.
Core capabilities
dsh-score compresses "can it be installed, is it maintained, are the docs complete, is it secure, does it comply" into one reviewable score card:
- Five dimensions — install success
25, maintenance20, documentation20, security20, compliance15(weights configurable; defaults sum to 100). - Evidence discipline — every dimension records its audit links (
source, sanitizeddetail,observedAt); a dimension without evidence reportsno-evidence(score 0, excluded from the weighted total) and never a fabricated number. - Structured results — every record carries
schema: "dsh-score/v1"with first-class fields; this is the machine-readable contract downstream tooling consumes. - Batch leaderboard —
/scorehands a target list to ascore-batchjob overctx.jobs, producing a JSON + Markdown leaderboard snapshot. - Cache and refresh — score cards are cached per target (24 hours by default);
refresh: truebypasses the cache and re-gathers evidence. - Embeddable badge —
score_badgereturns a shields.io badge (endpoint / SVG / Markdown) plus the five-dimension JSON, and the total badge drops straight into any README.
target ──gh api / npm view──▶ five dimensions of evidence ──weighted──▶ score card dsh-score/v1 ──▶ leaderboard lb_...
│
└─ no evidence → no-evidence (score 0, excluded)
Tools & results
Tool surface
| Entry | Type | Notes |
|---|---|---|
score |
tool | Scores one target through the five dimensions; refresh: true bypasses the cache, background: true starts a score-batch job |
/score <targets...> |
command | Scores a whitespace/comma-separated target list; progress streams through the job output and the final line names the lb_... for score_report |
score_report |
tool | Fetches a score card (sc_...), a leaderboard (lb_...), or — with no id — the latest leaderboard |
score_badge |
tool | Generates a shields.io badge and the five-dimension JSON for a target or a stored score card |
score-batch |
background job | The batch job over ctx.jobs; it scores serially and writes the leaderboard snapshot |
Tool parameters
score(target: string, refresh?: boolean, background?: boolean)
target— a GitHub repository (github:owner/repo,owner/repo, a git/https URL) or an npm package name.refresh: true— bypass the score cache and re-gather evidence.background: true— start ascore-batchjob and return its id.
score_report(id?) returns a score card (sc_...), a leaderboard (lb_...), or — with no id — the latest leaderboard.
score_badge(target? | id?, refresh?):
target— score a GitHub repo or npm package through the cache and badge it; mutually exclusive withid.id— badge a stored score card (sc_...) without re-scoring.refresh: true— bypass the score cache (only applies totarget).
Score card example
These are the plugin's real entry points; install once, then call one.
Command:
score({ target: "owner/repo" })
Output:
# Score sc_65e817eec8cd71ed
- Target: `owner/repo` (repo)
- Grade: **B** — total 88/100
- Verdict: healthy (weighted total 88/100)
| Dimension | Status | Score | Weight | Summary |
|---|---|---|---|---|
| Install success | ⬜ no-evidence | 0 | 25 | no dsh-test-drive result recorded for this target (install success unmeasured) |
| Maintenance | ✅ pass | 100 | 20 | active (2026-08-15T00:00:00.000Z; 0 open issues) |
| Documentation | 🟠 warn | 70 | 20 | 1/4 language READMEs + CHANGELOG + SECURITY |
| Security | ✅ pass | 100 | 20 | license MIT |
| Protocol compliance | ✅ pass | 81 | 15 | dsh.bundle.patch declared |
Observed from README.md:141-149,169-191; tests/tools.spec.ts:64-65; tests/score.spec.ts:50-55; src/report.ts:40-53
Failure branch — input:
{"id":"sc_nope"}
Output:
isError=true; error.message=dsh-score: no score recorded with id sc_nope
Observed from plugins/dsh-score/src/tools.ts:351-354 (throw); plugins/dsh-score/tests/tools.spec.ts:95-99 (unknown id); plugins/dsh-score/README.md:179-181 (score_report); exercised through plugins/dsh-score/tests/harness.ts
Brief examples: score({ target: "owner/dsh-click" }) · score_report("sc_65e817eec8cd71ed") · score_badge("owner/dsh-click")
Structured result
{
"schema": "dsh-score/v1",
"scoreId": "sc_8f1c2e4a9b3d7f01",
"target": { "kind": "repo", "spec": "github:owner/dsh-click#abc123" },
"scoredAt": "2026-08-16T00:00:00.000Z",
"durationMs": 3210,
"pluginVersion": "0.2.22",
"dimensions": {
"install": { "dimension": "install", "status": "no-evidence", "score": 0, "weight": 25,
"summary": "no dsh-test-drive result recorded for this target (install success unmeasured)",
"evidence": [{ "source": "test-drive", "detail": "no test-drive record found in the test_drive domain", "observedAt": "2026-08-16T00:00:00.000Z" }] },
"maintenance": { "dimension": "maintenance", "status": "pass", "score": 100, "weight": 20,
"summary": "active (2026-08-10T00:00:00Z; 0 open issues)",
"evidence": [{ "source": "gh-api", "detail": "last activity 2026-08-10T00:00:00Z", "observedAt": "2026-08-16T00:00:00.000Z" }] }
},
"total": 88,
"grade": "B",
"verdict": "healthy (weighted total 88/100)"
}
Scoring: the total is a weighted average over the dimensions that gathered evidence (no-evidence dimensions are excluded and renormalized); A ≥ 90, B ≥ 75, C ≥ 60, D ≥ 40, else F, and N/A when nothing had evidence.
Badge & five-dimension JSON
score_badge generates an embeddable README badge and the five-dimension JSON for one scored target. The badge comes in three forms derived from the same card: endpoint (a documented shields.io static URL, zero self-hosting), SVG (a self-contained shields.io flat-style SVG, badge.svg field / renderScoreBadge, for offline or self-hosted READMEs), and Markdown (the embed snippet combining both).
Embed the total badge:

The same call returns the compact JSON API envelope (schema: "dsh-score/badge/v1"):
{
"schema": "dsh-score/badge/v1",
"target": { "kind": "repo", "spec": "github:owner/dsh-click#abc123" },
"scoredAt": "2026-08-16T00:00:00.000Z",
"total": 84,
"grade": "B",
"dimensions": {
"install": { "label": "install", "status": "no-evidence", "score": 0, "weight": 25, "summary": "no dsh-test-drive result recorded" },
"maintenance": { "label": "maintenance", "status": "pass", "score": 90, "weight": 20, "summary": "active (0 open issues)" },
"documentation": { "label": "docs", "status": "pass", "score": 85, "weight": 20, "summary": "README + CHANGELOG + SECURITY" },
"security": { "label": "security", "status": "warn", "score": 60, "weight": 20, "summary": "permissive license" },
"compliance": { "label": "compliance", "status": "pass", "score": 100, "weight": 15, "summary": "dsh.bundle.patch + dsh-plugin topic" }
}
}
A no-evidence dimension keeps its honest status and score 0 — the badge and JSON never fabricate a number.
Configuration
All keys are optional (defaults shown); invalid values fail loudly at load.
Most-used keys:
| Key | Default | Meaning |
|---|---|---|
probeTimeoutMs |
60000 |
Deadline for one gh/npm probe command (ms) |
outputTailBytes |
8000 |
Cap on the sanitized output tail recorded per probe (bytes) |
cacheMaxAgeMs |
86400000 |
How long a cached score card is reused before re-scoring (0 disables the cache) |
staleCommitWarnDays |
90 |
Commit/publish age at which maintenance drops to warn |
staleCommitFailDays |
365 |
Commit/publish age at which maintenance drops to fail |
maxBatchTargets |
20 |
/score batch target cap |
batchConcurrency |
1 |
Batch concurrency (serial by default, avoiding API-rate contention) |
weights |
{install:25, maintenance:20, documentation:20, security:20, compliance:15} |
Per-dimension weights (each 0–100; at least one must be > 0) |
| Key | Default | Meaning |
|---|---|---|
probeTimeoutMs |
60000 |
Deadline for one gh/npm probe command (ms) |
outputTailBytes |
8000 |
Cap on the sanitized output tail recorded per probe (bytes) |
cacheMaxAgeMs |
86400000 |
How long a cached score card is reused before re-scoring (0 disables the cache) |
staleCommitWarnDays |
90 |
Commit/publish age at which maintenance drops to warn |
staleCommitFailDays |
365 |
Commit/publish age at which maintenance drops to fail |
staleIssueWarnDays |
30 |
Oldest-open-issue age (response proxy) at which maintenance drops to warn |
staleIssueFailDays |
180 |
Oldest-open-issue age at which maintenance drops to fail |
maxBatchTargets |
20 |
/score batch target cap |
batchConcurrency |
1 |
Batch concurrency (serial by default, avoiding API-rate contention) |
weights |
{install:25, maintenance:20, documentation:20, security:20, compliance:15} |
Per-dimension weights (each 0–100; at least one must be > 0) |
Example override in your profile patch:
- insert:
- id: score
name: dsh-score
config:
cacheMaxAgeMs: 0
maxBatchTargets: 10
weights:
install: 25
maintenance: 20
documentation: 20
security: 20
compliance: 15
FAQ
Score cards and leaderboards are stored in the score storage domain (tables scores, leaderboards, plus the latest-leaderboard pointer). When the composition has no storageDomain (a headless profile older than the base bundle's storage rows, for example), the tools still work but score persistence is disabled with a logged reason. The published dsh-base bundle mounts storage-domain from 0.1.2-rc.1 on (verified against the 0.1.2-rc.1 and 0.1.5-alpha.1 tarballs), so persistence is active on the published line.
No. A failed probe, unparsable output, or a target with no resolvable GitHub repository makes that dimension report no-evidence (score 0, excluded from the weighted total and renormalized); when nothing has evidence the total is N/A. A no-evidence dimension also carries the failure reason (exit N, timeout, the redacted stderr tail, or invalid JSON) instead of an invented number.
No. The pipeline runs only the read-only gh api and npm view commands — it never installs, builds, or executes the target. Every CLI invocation is an argv array, never shell-interpreted, and the owner/repo segments are validated against a restricted character set before they reach an endpoint.
No. Child processes inherit the provider's credential-scrubbed environment and gh reads its own credential store; no environment value is ever logged. All report/log strings pass through pure sanitizers: token literals, URL credentials, and bearer headers are redacted, and output tails are byte-capped.
Score cards are cached per target (cacheMaxAgeMs = 24 hours by default), so repeat lookups do not re-gather evidence; use refresh: true or set the cache age to 0 when you need fresh data. /score defaults to batchConcurrency: 1 (serial) to avoid concurrent probes competing for API quota, and a single probe waits at most probeTimeoutMs (60 seconds by default).
Security boundaries
- No code execution. The pipeline runs
gh apiandnpm viewonly; it never installs, builds, or runs a target. - Argv-only subprocesses. Every CLI invocation is an argv array, never shell-interpreted; repo owner/repo segments are validated against a restricted character set before use in an endpoint.
- Evidence discipline. No score is fabricated: a probe that fails or returns unparsable output yields
no-evidence, never a number. - Detection vs redaction. Secret-leak and malicious-install-script detection share the same pure regexes as redaction; both are unit-tested against extreme inputs.
Permissions & data
- Permissions: only public services are consumed —
ctx.subprocess,ctx.jobs,ctx.storageDomain,ctx.tools,ctx.commands; the plugin provides no Cordis service key. - Data: score cards and leaderboards live in the
scorestorage domain (tablesscores,leaderboards, plus the latest-leaderboard pointer). WithoutstorageDomainthe tools still work and persistence is disabled with a logged reason. - Environment: child processes inherit the provider's credential-scrubbed environment;
ghreads its own credential store. No environment value is ever logged, and every report/log string passes through pure sanitizers with byte-capped tails.
Known limitations
- Repository probes require
ghto be authenticated and network access to GitHub; npm probes requirenpmand registry access. - A target without a resolvable GitHub repository cannot be inspected for documentation, security, or compliance (those dimensions report
no-evidence). - Install success depends on
dsh-test-drivebeing mounted and having recorded the target; otherwise it is honestlyno-evidence. - The maintenance "issue response" signal is a proxy (oldest open issue age), not a direct response-time measurement.
- Score results are cached per target; use
refresh: true(or wait pastcacheMaxAgeMs) to force re-scoring.
Interoperability with other DSH plugins
dsh-score only exposes its own scoring surface and never takes over another plugin's lane:
- Self-contained tool surface. It registers exactly three tools (
score,score_report,score_badge) and one command (/score), all throughctx.effect, so stop/hot-reload removes them. - No service key.
injectdeclares onlytools,commands,subprocess, andjobs;storageDomainis read optionally throughctx.getand never required. - No client lane. It registers no client slot key and no
webServerroute prefix. - No patch-layer collision. The bundle patch only
inserts its own row (id: score); it never overrides a built-in row'sconfig. - No global mutation. It does not patch prototypes, rewrite
process.env, or replace the global fetch dispatcher. - No shared-event listeners. It registers no waterfall listener (including
agent/pre-step), so it takes no part in event-chain ordering and cannot short-circuit another plugin's chain.
Static evidence: the dsh-plugin-doctor verification badge is driven by this repository's CI output (see "Ecosystem & quality").
Compatibility
| Component | Version |
|---|---|
| DeepSeek Harness | dsh-v0.2.2-rc.4 (GitHub tag). The engines.dsh gate is >=0.2.1-0 <0.3.0 || >=0.2.2-0 <0.3.0; dev/test pins and the typecheck:ci ruler measure 0.2.2-rc.1. The ctx.jobs seam uses the SessionId contract. |
| Node.js | ^22.19.0 || >=24.9.0 |
| Package manager | pnpm@11.7.0 |
| Platform | Windows / macOS / Linux (host-only plugin) |
| External tools | gh CLI on PATH (authenticated for API reads), npm CLI on PATH |
Ecosystem & quality
- Gitee mirror (sync mirror; GitHub is canonical)
- dshfind plugin page (installs and ranking)
- OpenSSF Scorecard report (driven by .github/workflows/scorecard.yml)
- dsh-plugin-doctor verification report (badge source)
- DSH Market (listed)
- 1024 store channel:
npm i -g dsh1024once, thendsh1024 plugin --profile web add dsh-score(counts toward the deepseek1024.com install ranking) - Ecosystem knowledge base — measured data, not marketing: plugin development guide · plugin-selection data · maintenance criteria
Development
pnpm install # node ^22.19 || >=24.9
pnpm run typecheck && pnpm run typecheck:ci && pnpm test # local checkout face + published-line face + vitest
pnpm run build && pnpm run verify:self-contained && pnpm run verify:artifacts && pnpm pack
pnpm run check:readmes # five-language README consistency
typecheckresolves@deepseek-ai/*through the local harness checkout;typecheck:cichecks against the published0.2.2-rc.1types.- Tests use the real
Context/Session/ToolRuntime/LocalJobRegistry/storage stack with a scripted subprocess provider. - Real-CLI scoring (requires
gh/npmon PATH andghauthenticated): invokescorefrom a mounted profile. - Release:
node scripts/release.mjs <x.y.z>(bumps, stamps CHANGELOG, re-runs the gate, commits + tags; never pushes). - The README images are generated programmatically by
scripts/gen-readme-assets.py(Pillow for the hero banner, hand-written SVG for the structure diagram) and are safe to re-run; the older single-skin flow diagram and demo images stay indocs/assets/for external deep links, while the READMEs reference only the new hero and flow-v5 structure diagram, each in light and dark.
Topics: dsh, dsh-plugin, deepseek-harness, deepseek, cordis, plugin-scoring, quality-score, leaderboard, supply-chain
Family & thanks
This project is part of the DeepSeek Harness plugin family from PerryLink (all Apache-2.0). If it is useful, a star helps the next person find it — nothing is gated behind it.
- @PerryLink — creator and maintainer: the five-dimension pipeline, the
gh/npmevidence probes, the structureddsh-score/v1contract, the/scorebatch job, the badge, and the five-language docs.
HCB compatibility input
The host compatibility matrix and the ENTRY/FLOW/BOOST tracks are published in HCB; this README does not maintain a second host-version matrix.
License
LICENSE (Apache License 2.0) © 2026 dsh-score contributors
Links
More in this category
yjh051108/dsh-routing-suite★ 7028
One repository, three parts: a runtime injector for DSH plugin packages (inject, hot-reload, unload, promote a dev staging tool to the front, route self-heal, plus a settings-page plugin manager that lists, unloads and drags folders in to internalize), a task-aware reasoning-mode router agent preset (router-standard / router-spec / router-react), and a graded two-level task protocol whose six tools (commit_star, lock_stage, revise_do, edit_plan, mark_task, redteam_verdict) pin task state to disk. The injector implementation ships in-tree, so the install carries its own behaviour rather than a dependency list.
strukto-ai/mirage#dsh★ 3798
Swaps the filesystem and bash providers for a mirage virtual workspace: file tools and shell commands run over mounted resources (RAM, S3, Redis, Slack, Gmail, Notion, Postgres) instead of the host disk, with per-mount read/write/exec modes, per-command sandbox routing (monty, pyodide, quickjs in process; docker, e2b, daytona remote), and installed CLIs (git, gh, slack, linear, ntn, gws, or one you register) as head words in the virtual terminal.
first-fluke/oh-my-agent#dsh★ 1473
Runs project OMA checks before and after tool execution and when a turn tries to finish. Installs its OMA CLI dependency and initializes unconfigured projects on first use. Requires Node.js 22+; initial setup needs network access and project write access.
MichengAI/dsh-skills-manager★ 598
Manages local DeepSeek Harness skills in Settings and shows shared Agent skills as read-only.
hust-open-atom-club/oh-dsh★ 324
Community distribution: TUI, desktop, and Web UI as one bundle with layered installation.
Fishquito7/dsh-skill-mcp-panel★ 287
Manages DSH skills and MCP servers from the web settings: skill cards with hot enable/disable, workspace scopes, groups, batch migration and drag-and-drop import, plus stdio/HTTP MCP CRUD with connection tests, secret redaction and the unified dsh-panel CLI.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.