Paperclip Dev Workspace Run Verify Fix Agent Skill

Paperclip Dev Workspace Run / Verify / Fix

This skill is for Paperclip-specific development workspaces whose service is

started through project execution workspace runtime services, typically a

worktree service such as paperclip-dev.

Success means all of these are true:

• the service was started through the normal managed runtime path, not a

detached workaround

• the worktree database is a full bootstrapped isolated clone of the primary

instance database

• /api/health returns status: ok and bootstrapStatus: ready
• the root page returns 200 and does not show the first-admin setup gate
• the board user can log in with their normal dev credentials
• the app shows populated cloned data, not only a manually copied auth user
• the main control plane shows the service as running / healthy with the

expected URL

• the served workspace app also knows about that service and shows it as

running / healthy

• the process that owns the service port really belongs to the target

workspace: readlink /proc/<pid>/cwd resolves inside the target worktree,

not a sibling workspace

If any item fails, keep fixing. Do not mark the issue done because one probe

passed.

Hard rules

• The master checkout /srv/paperclip/home/paperclipai/paperclip (the primary

repo with the real instance .env) is never a workspace repair target. Never

point a worktree runtime service at it, never git worktree add into it or

under <master>/.paperclip/worktrees/, and never edit its .env while

repairing a workspace. Managed workspaces run in their own folders, for

example ~/.paperclip-worktrees/instances/<slug>/….

• Never trust a running / healthy runtime row or a passing /api/health

alone. The service:paperclip-dev port is pinned in service config, so a

sibling workspace's process can own the port and still answer health checks.

Confirm the port owner's real cwd (readlink /proc/<pid>/cwd) before and

after every repair; the runtime row's recorded cwd and pid can be

fabricated by port adoption.

• Read doc/DEVELOPING.md before running Paperclip CLI, dev server, worktree,

database, build, or test commands.

• Use the Paperclip CLI and API as the source of truth for worktree and

database operations. Do not use psql, raw embedded-Postgres commands, or

ad hoc row copying for the normal fix path.

• Do not manually copy only auth rows as the final fix. That proves login, but

it does not prove the isolated workspace has the full bootstrapped database.

• Prefer managed runtime start/stop routes over pnpm dev or detached shell

processes when the task is about a reusable workspace service.

• Avoid destructive git or database actions. Preserve user changes in the

worktree.

• Run the smallest verification that proves the repair. Add focused tests only

when code changed.

• Leave an issue comment with root cause, exact fix, verification, and any

commit link. Set a clear final status.

Inputs to collect

Use environment variables when available. Do not print API keys or passwords.

• PAPERCLIP_API_URL: main control-plane API URL
• PAPERCLIP_API_KEY: agent API key
• PAPERCLIP_RUN_ID: current run id for runtime-service mutations
• PAPERCLIP_TASK_ID: current issue id
• PAPERCLIP_COMPANY_ID: company id
• PAPERCLIP_AGENT_ID: current agent id
• execution workspace id for the worktree service
• runtime workspace command id, usually service:paperclip-dev
• expected service URL, for example http://paperclip-dev:40631
• dev credential owner, if the user supplied one; never post the password

If the execution workspace id or service command id is missing, read the issue,

project, or execution-workspace API records first. Do not guess and start an

unmanaged server on a random port.

Normal run sequence

Use this when the user says to start the workspace, start it again, or fix a

workspace that should be freshly ready.

• Confirm the latest issue comment and restate the success condition in your

own words.

• Inspect current runtime state from the main control plane.
• Check the target port for conflicts: identify the current port owner and

its real cwd, and check whether any sibling workspace's runtime row also

claims the same port (see "Port conflicts and workspace identity").

• If another live workspace or agent run owns the port, stop that owner

through its managed control path and verify the port stays free. Do this

before restarting the target; otherwise the owner can respawn and reclaim

the port.

• Stop any managed instance of the target runtime service.
• Reseed the worktree with a full clone from the primary instance when there

is any doubt about database completeness.

• Start the runtime service through the managed runtime API.
• Verify health, bootstrap state, login readiness, populated data, runtime

visibility, and port-owner identity from both the main control plane and

served workspace app.

• If a verification item fails, diagnose that exact failure and loop back to

the narrowest repair step.

10. Comment on the issue and set the final disposition.

Managed start / stop

Use the runtime-service endpoints on the main control plane. Include

X-Paperclip-Run-Id so the mutation is associated with the current heartbeat.

curl -sS -X POST \
  "$PAPERCLIP_API_URL/api/execution-workspaces/$EXECUTION_WORKSPACE_ID/runtime-services/stop" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" \
  -H "X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID" \
  -H "Content-Type: application/json" \
  --data-binary '{"workspaceCommandId":"service:paperclip-dev"}'

curl -sS -X POST \
  "$PAPERCLIP_API_URL/api/execution-workspaces/$EXECUTION_WORKSPACE_ID/runtime-services/start" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" \
  -H "X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID" \
  -H "Content-Type: application/json" \
  --data-binary '{"workspaceCommandId":"service:paperclip-dev"}'

If the API returns an existing service, treat that as a candidate only. Verify

its real /api/health and its port-owner identity before trusting it.

Port conflicts and workspace identity

service:paperclip-dev pins an explicit port in service config, so every

workspace that realizes that service claims the same port. On start, service

adoption will adopt whatever process already owns the port when its command

line looks similar — and it records the *requested* cwd, not the owner's real

cwd. Two consequences:

• multiple workspaces can hold running rows for the same port while only one

process exists; whichever workspace URL you open, you get that one process

• the runtime row's cwd, pid, and health can all look correct while the

URL actually serves a sibling worktree's app

Identity check — run this before trusting any existing service and again after

every start or restart:

pid=$(lsof -nP -iTCP:"$SERVICE_PORT" -sTCP:LISTEN -t | head -1)
readlink "/proc/$pid/cwd"

The resolved cwd must be inside the target worktree (the dev server usually

runs in <worktree>/server). Only /proc/<pid>/cwd is authoritative. Do not

accept the runtime row's cwd field, a root 200, or a healthy /api/health

as proof of identity — a squatting sibling passes all three.

If the port owner is a sibling workspace's process (port squat):

• Decide which workspace should own the pinned port right now — normally the

workspace named in the issue.

• Identify the current owner's workspace and whether a live agent run or

managed runtime service is responsible for keeping it alive. Use the real

/proc/<pid>/cwd, the sibling execution-workspace records, and the owning

issue/run state; do not infer ownership from the target runtime row.

• Stop the owner first through its own managed control path: stop the sibling

workspace's runtime service, or stop/cancel the live run that supervises and

respawns it. Do not kill the sibling process directly while its supervisor

is still active, because it will simply come back and reclaim the port.

• Verify the port remains unbound for a stability wait. If it is immediately

rebound, find and stop the still-live supervisor instead of restarting the

target in a loop.

• Only after the conflicting owner is stopped, start or restart the target

service and re-run the identity check:

curl -sS -X POST \
  "$PAPERCLIP_API_URL/api/execution-workspaces/$EXECUTION_WORKSPACE_ID/runtime-services/restart" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" \
  -H "X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID" \
  -H "Content-Type: application/json" \
  --data-binary '{"workspaceCommandId":"service:paperclip-dev"}'

• A restart only holds if the conflicting owner stays stopped. If the port is

reclaimed, stop and diagnose the owner/supervisor again; do not loop target

restarts against it.

• If two workspaces genuinely need to run at the same time, they cannot share

a pinned port. Stop the one that is not needed (through its own workspace's

managed stop, or by ending the run that keeps respawning it), say in your

issue comment which workspace now owns the port, and escalate the

pinned-port collision as a product issue instead of looping restarts.

Before declaring success, also fetch the sibling workspaces you suspect and

compare their runtimeServices[].port values: more than one running row

claiming the target port across workspaces is always a conflict to resolve,

not a state to work around.

Full database reseed

Use a full reseed when the app says setup is incomplete, login works but data

is missing, the cloned app does not have the expected companies/issues/agents,

or the user explicitly asks for the normal isolated-workspace database.

npx paperclipai worktree reseed --from-instance default --seed-mode full --yes

After reseed, restart through the managed runtime path. A reseed can copy

runtime-service rows whose ids no longer match the local process registry, so

runtime adoption and reconciliation must be verified after the start.

Do not consider the reseed complete until the served app has both auth and

populated product data. A one-off inserted user/account row is a diagnostic

clue, not the final state.

Core verification probes

Set SERVICE_URL to the service URL returned by the runtime API.

curl -sS "$SERVICE_URL/api/health" | jq
curl -sS -I "$SERVICE_URL/" | head

Expected health:

• status: "ok"
• bootstrapStatus: "ready"
• bootstrapInviteActive: false

Failures to reject:

• bootstrap_pending: the instance will show the first-admin setup gate
• database_unreachable: a web process is listening but its database is dead
• a root 200 with unhealthy /api/health: stale process adoption bug or a

dead embedded database behind a live Node process

Check the port owner when a process is already listening:

lsof -nP -iTCP:"$SERVICE_PORT" -sTCP:LISTEN || true

Use this only to identify and remove a stale matching Paperclip dev-runner

process after managed stop fails. Do not kill unrelated processes.

Verify main control-plane runtime state

Read the execution workspace from the main API and inspect the runtime service

record.

curl -sS \
  "$PAPERCLIP_API_URL/api/execution-workspaces/$EXECUTION_WORKSPACE_ID" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" | jq

The target service should show:

• matching workspace command id, usually service:paperclip-dev
• status: "running"
• healthy health fields, if present
• the same URL you are probing
• a local provider reference that maps to the running port owner

If the main app says running but /api/health is bad, stop and replace the

stale process through the managed runtime.

Verify served workspace runtime state

The cloned Paperclip app must also know about the service. Query the same

execution workspace through the served app when agent auth is available there:

curl -sS \
  "$SERVICE_URL/api/execution-workspaces/$EXECUTION_WORKSPACE_ID" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" | jq

The served app should agree that the service is running / healthy at the

same URL. If the main control plane and served app disagree after a reseed,

the cloned database may contain copied runtime-service ids that do not match

the local process registry. Use the normal start/adoption path again and verify

both sides. If code changed in this area, add a focused regression test.

Verify auth and populated data

Auth and data checks should use product APIs and browser/QA review, not raw DB

queries.

Minimum API checks:

curl -sS "$SERVICE_URL/api/health" | jq '.status, .bootstrapStatus'
curl -sS "$SERVICE_URL/api/companies" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" | jq
curl -sS "$SERVICE_URL/api/agents/me" \
  -H "Authorization: Bearer $PAPERCLIP_API_KEY" | jq

Then verify at least one expected cloned product record through the API, such

as a known project, issue key, company, or execution workspace that should

exist in the primary instance. Pick a record relevant to the current issue

rather than a random table count.

Browser or QA check:

• open the service URL
• confirm the first-admin setup gate is gone
• sign in with normal dev credentials
• confirm the board loads populated companies/projects/issues/runs
• confirm the workspace service appears with the running URL

If this environment cannot launch a browser, ask QA to do the visual/login

check and still complete all API checks you can run. Report that browser

verification was delegated and why.

Common failures and fixes

Service URL serves a different workspace (port squat)

Symptom: the workspace URL loads a working Paperclip app, but it is a sibling

worktree's app — wrong branch, wrong data, or the UI keeps bouncing you into

another workspace's pages.

Likely cause: the pinned service port is owned by another workspace's process

and start-time adoption attached that process to this workspace's runtime row.

Health checks pass because the sibling app is genuinely healthy.

Fix: follow the owner-first procedure in "Port conflicts and workspace

identity": identify and stop the sibling service or supervising run, verify

the port stays free, then start or restart the target service. Re-verify that

/proc/<pid>/cwd of the new port owner resolves inside the target worktree.

Verify: identity check passes, and one workspace-specific record served

through the URL (branch, issue key, or workspace id) matches the target

workspace.

Recorded port drifts after an ad hoc start

Symptom: the runtime row and the actual listener disagree about the port, or a

previously pinned port changed after someone started the app by hand.

Likely cause: an unmanaged pnpm dev (for example with --bind lan) picks its

own port instead of using the managed pinned port.

Fix: stop the unmanaged process, then managed start. Verify the recorded URL

port equals the bound port and the identity check passes.

Setup gate appears

Symptom: the page says no admin has claimed the instance.

Fix: run a full worktree reseed from the primary instance, then restart the

managed service. Claiming a first admin can clear the gate, but if the user

asked for the normal isolated workspace database, full reseed is the correct

fix.

Verify: /api/health has bootstrapStatus: ready, login works, and populated

data exists.

Login fails

Symptom: bootstrap is ready, but the user's normal dev credentials do not work.

Likely cause: the isolated DB has roles or bootstrap state but lacks the

primary instance auth users/accounts.

Fix: full reseed. Do not manually copy only Better Auth user/account rows as

the final fix.

Verify: user login through browser/QA and /api/agents/me with the agent key.

Login works but data is missing

Symptom: user can sign in, but companies/issues/projects/runs are empty or

clearly incomplete.

Likely cause: a partial auth repair was done instead of a full cloned database.

Fix: full reseed, managed restart, then verify representative cloned records.

Port listens but health says database unreachable

Symptom: curl -I / returns a response, but /api/health reports

database_unreachable.

Likely cause: stale Node/web process remained alive after embedded Postgres

died.

Fix: managed stop first. If the process survives, identify the matching

Paperclip dev-runner process group for the target port and terminate only that

group. Then managed start.

Verify: /api/health is ok after a stability wait and the runtime record is

healthy.

Main control plane loses track of a running service

Symptom: the service URL works, but the main app says the service was not

created or is stopped.

Likely cause: detached workaround process, stale provider ref, or service

adoption trusted the root URL instead of health.

Fix: shut down the unmanaged process and restart through the managed runtime.

If code repair is needed, ensure adoption checks /api/health, replaces

unhealthy adopted processes, and records the current provider ref.

Verify: main runtime row and /api/health agree.

Served app loses track after full reseed

Symptom: the main app sees paperclip-dev running, but the cloned app copied a

runtime-service row whose id does not match the local registry.

Likely cause: normal DB clone copied persisted runtime rows from the primary

instance into an isolated environment with different local process metadata.

Fix: use managed start/adoption again. If code repair is needed, adoption

should reconcile by service identity and port, not only by copied row id.

Verify: main app and served app both show the same service as

running / healthy.

Runtime start returns permission or run-FK errors in the cloned app

Symptom: starting from the served app returns 403 or a mutation partially

applies before activity logging fails.

Likely causes: the cloned issue/run/agent state does not match the current

heartbeat, or the run id is absent in the cloned database after reseed.

Fix: prefer the main control-plane managed runtime path and full reseed. If

the cloned app state itself must be repaired, use normal Paperclip issue/run

transitions first. Do not hide the condition with raw DB edits; report the

exact guard or missing row if it blocks the normal path.

Recurring gotchas from past repairs

Each of these has burned a previous repair; check them before deep debugging.

• Worktree instance logs live at

~/.paperclip-worktrees/instances/<slug>/logs. Read them before guessing at

login or startup failures.

• Probing the service through a LAN or tailscale IP can return 403 while

127.0.0.1 (or the paperclip-dev hostname) works. Probe loopback first

before concluding auth is broken.

• Browser QA that lands on a blank company path (for example /FOR/...)

should switch to a company key that exists in the cloned data (for example

/PAP/...) before concluding the reseed failed.

• A 409 on runtime or issue mutations usually means another live run holds

the ownership lock. Do not force it; wait or coordinate through the issue.

• After a full reseed the cloned DB can contain copied runtime-service rows

that do not match local processes. Re-run managed start and the identity

check instead of trusting copied rows.

When code changes are required

Make a code change only when the normal operational repair exposes a product

bug. Examples from this failure class:

• local port owner detection used the wrong lsof arguments
• service adoption trusted root 200 instead of /api/health
• port-owner adoption recorded the requested cwd instead of verifying the

owner's real /proc/<pid>/cwd, so a sibling workspace's process was adopted

• pinned service ports collided across sibling workspaces
• unhealthy adopted services were not terminated/replaced
• reseeded runtime-service ids were not reconciled with local process registry

Add focused tests in the affected service test file. For workspace runtime

repairs, the narrow verification is usually:

pnpm exec vitest run server/src/__tests__/workspace-runtime.test.ts
git diff --check

Commit logical code changes and link the commit in the issue comment. If no

tracked code changed, say so explicitly.

Final issue comment template

Use concrete evidence, not a vague "it works".

Fixed and verified the workspace service.

Root cause:
- <why it broke>

Fix:
- <normal reseed/start/repair steps>
- <code commit if any>

Verified:
- main control plane shows <service> running/healthy at <url>
- served workspace app shows the same service running/healthy
- port owner identity: /proc/<pid>/cwd resolves inside the target worktree
- <url>/api/health is ok with bootstrapStatus ready
- root page returns 200 and no setup gate
- dev login verified by <agent browser / QA / user> without posting credentials
- cloned data verified via <specific API records>
- targeted tests: <commands>

Remaining:
- <none, or named owner/action if blocked>

Mark the issue done only when every success-condition item is satisfied. If

not, mark blocked with a named unblock owner and the exact action needed.