CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

30.8 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 1 flagged

Securityscore 88 · 1 finding

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · README.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptREADME.md· markdown

342## Accounts and credentials

343

344SC should never ask you to paste a password, API key, or access token into chat or machine-t

… (9 chars elided on L344)

345

346When account access is required, SC should tell you:

Occurrences

1 occurrence · at L344

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25614b7852251bf4666…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

skills/sc-dokploy/SKILL.md · 1 file

skills/sc-dokploy/SKILL.md8.1 KB · Markdown

name: sc-dokploy

description: "Dokploy CRUD, audit, and debug. List/create/update/delete projects, applications, compose services, and domains via REST API. Find stale domains, duplicate hosts, and *.traefik.me leftovers. Inspect status and recent deployments."

use_when: "Use when the task matches this skill scope: Dokploy CRUD, audit, and debug. List/create/update/delete projects, applications, compose services, and domains via REST API. Find stale domains, duplicate hosts, and *.traefik.me leftovers. Inspect status and recent deployments."

do_not_use_when: "Do not use when the task is outside this skill scope or a more specific SI-Coder skill owns the requested outcome."

required_tools: []

security_constraints: "Never request, print, or persist plaintext credentials in chat/tool payloads; use SI-Coder safe credential handoffs."

references: []

compatibility: "Standalone SI-Coder; host invocation syntax and available tools may vary."

/sc-dokploy — Dokploy CRUD & Audit

Use this skill when the user wants to inspect, change, or clean up Dokploy state directly (without redeploying code).

Pre-requisites

DOKPLOY_API_URL, DOKPLOY_API_KEY — Dokploy admin
SSH fallback to the Dokploy host (for orphan swarm services / Traefik file CRUD the REST API does NOT expose): ssh -i ~/.ssh/<your_key> <user>@<your-dokploy-host> with passwordless sudo.

If missing, route to /sc-onboarding.

Connection-safe local execution

When Dokploy access is stored in an SI-Coder named source=sc connection, run credential-dependent commands through sc run -- ... so the selected connection is injected only into that child process:

sc run -- node skills/sc-dokploy/scripts/projects.js list

The shorter node scripts/... examples below describe the script interface and are valid only when the required environment is already supplied by the caller. Do not export Dokploy credentials globally just to use them. External/provider-managed sources must execute through their own connection backend instead of sc run.

REST vs SSH

Dokploy REST API covers projects, applications, compose, domains, deploy/start/stop, monitoring read. It does NOT cover:

docker service rm (orphan swarm services)
rm /etc/dokploy/traefik/dynamic/<file>.yml (orphan Traefik routers)
container exec / kill / log tail beyond the dashboard

For those, SSH in directly. Useful one-liners:

ssh -i ~/.ssh/<your_key> <user>@<your-dokploy-host> 'sudo -n docker service ls'
ssh -i ~/.ssh/<your_key> <user>@<your-dokploy-host> 'sudo -n ls /etc/dokploy/traefik/dynamic/'
ssh -i ~/.ssh/<your_key> <user>@<your-dokploy-host> 'sudo -n docker service rm <name>'
ssh -i ~/.ssh/<your_key> <user>@<your-dokploy-host> 'sudo -n rm /etc/dokploy/traefik/dynamic/<name>.yml'  # Traefik file watcher reloads in ~5-10s

Orphan-service pattern: when a Dokploy app is recreated, the old swarm service and its Traefik dynamic config can survive deletion. Both compete for the same Host(...) rule. Symptom: prod serves an old image even after a fresh deploy. Diagnosis: docker service ls shows two services for the same project, and two .yml files in traefik/dynamic/ bind the same domain. Fix: remove the orphan file + service.

CORE RULES

Idempotency: after a duplicate error, inspect the existing domain and confirm host, service and port match. Only that verified duplicate is a no-op; do not swallow other 4xx errors.
Don't delete domains blindly: only delete domains via audit.js's selectDomainsToDelete — keep the desired canonical host, drop *.traefik.me and duplicates.
Never rename the Dokploy control host: whatever hostname is in DOKPLOY_API_URL is the control plane. Never rewrite it inside scripts — read it from env.
`x-api-key` header, not Bearer: Dokploy uses x-api-key, NOT Authorization: Bearer.

Release source and evidence

Prefer the repository's release function over generic deploy calls. Inspect Dokploy source mode: Git and drop-source releases have different contracts. For drop source, Git push alone does not update the source directory. Publish a clean archive of the verified commit, preserve server-managed env, project the revision metadata and queue one rollout. Serialize concurrent agent deployments. Require terminal rollout status and live health matching that commit. Follow the main sc delivery workflow for domain/auth/email and CI/CD acceptance.

Scripts

`projects.js` — Project CRUD

node scripts/projects.js list
node scripts/projects.js create <name>
node scripts/projects.js show <name>

`apps.js` — Application CRUD

node scripts/apps.js list --project <name>
node scripts/apps.js show --project <name> --app <name>
node scripts/apps.js deploy --project <name> --app <name>

`compose.js` — Compose service CRUD

node scripts/compose.js list --project <name>
node scripts/compose.js show --compose <composeName>
node scripts/compose.js env --compose <composeName>
node scripts/compose.js deploy --compose <composeName>

`domains.js` — Domain CRUD

node scripts/domains.js list-app --app-id <id>
node scripts/domains.js list-compose --compose-id <id>
node scripts/domains.js create-app --app-id <id> --host <host> [--port N]
node scripts/domains.js create-compose --compose-id <id> --host <host> --port <n> --service <name>
node scripts/domains.js delete --domain-id <id>

`audit.js` — Sweep

Reports across all projects:

*.traefik.me placeholder hosts that should be removed
Duplicate hosts on the same service
Applications with no domain configured
Compose services missing INSTANCE_SECRET

node scripts/audit.js [--fix]   # --fix removes stale domains

`debug.js` — Status & recent deployments

node scripts/debug.js status --project <name> --app <name>
node scripts/debug.js status --compose <composeName>
node scripts/debug.js deployments --app-id <id>

API endpoint reference (Dokploy)

Action	Endpoint	Method
List projects	`/project.all`	GET
Create project	`/project.create`	POST `{ name }`
Get application	`/application.one?applicationId=`	GET
Update application	`/application.update`	POST
Deploy application	`/application.deploy`	POST `{ applicationId }`
Get compose	`/compose.one?composeId=`	GET
Update compose	`/compose.update`	POST
Deploy compose	`/compose.deploy`	POST `{ composeId }`
Deploy compose template	`/compose.deployTemplate`	POST `{ environmentId, id }`
Create domain	`/domain.create`	POST
Delete domain	`/domain.delete`	POST `{ domainId }`
List GitHub providers	`/github.githubProviders`	GET
Save app GH provider	`/application.saveGithubProvider`	POST

Auth: x-api-key: <DOKPLOY_API_KEY>.

Note on logs

Dokploy build logs are NOT exposed over the REST API (as of this skill's last update). On deployment failure, point the user at the Dokploy dashboard: <DOKPLOY_API_URL without /api> → project → service → Deployments

Existing domain vs public-IP fallback

For a port-exposed managed service or webapp that can run without DNS, domain configuration is an enhancement, not an install dependency:

If the deployment/app already has a canonical domain and its DNS/TLS route is configured, preserve and prefer that domain for browser/UI access. Never replace a working existing domain with a newly invented subdomain.
If no domain is configured, do not block a healthy install solely for DNS. Keep the declared public port reachable on the server public IP when the app's security model allows it, and report http://<public-ip>:<port> as the fallback.
When both exist, the domain is the primary UI/embedded origin; public IP is fallback/diagnostic only.
Domain provider automation runs only for an explicit desired domain or an existing project domain that needs repair. Missing Hostinger/Cloudflare credentials must not cause an unrelated port-based runtime install to fail.
Verify an existing-domain conflict against the intended host/service/port before treating it as a no-op. Do not delete or recreate a working route merely to make automation own it.

sc-dokploy— agent skill

Is sc-dokploy safe to install?

/sc-dokploy — Dokploy CRUD & Audit

Pre-requisites

Connection-safe local execution

REST vs SSH

CORE RULES

Release source and evidence

Scripts

`projects.js` — Project CRUD

`apps.js` — Application CRUD

`compose.js` — Compose service CRUD

`domains.js` — Domain CRUD

`audit.js` — Sweep

`debug.js` — Status & recent deployments

API endpoint reference (Dokploy)

Note on logs

Existing domain vs public-IP fallback

Audit the pieces. Scan the whole. Decide.

Is sc-dokploy safe to install?

/sc-dokploy — Dokploy CRUD & Audit

Pre-requisites

Connection-safe local execution

REST vs SSH

CORE RULES

Release source and evidence

Scripts

projects.js — Project CRUD

apps.js — Application CRUD

compose.js — Compose service CRUD

domains.js — Domain CRUD

audit.js — Sweep

debug.js — Status & recent deployments

API endpoint reference (Dokploy)

Note on logs

Existing domain vs public-IP fallback

Audit the pieces. Scan the whole. Decide.

`projects.js` — Project CRUD

`apps.js` — Application CRUD

`compose.js` — Compose service CRUD

`domains.js` — Domain CRUD

`audit.js` — Sweep

`debug.js` — Status & recent deployments