Troubleshooting
Find your error below and follow the next step.
The console
A waitlist page still appears. Core skill sharing no longer requires activation. Reload the console to get the current release. If it persists, contact Slancha.
"requires the 'approver' role; you have 'viewer'." Ask a workspace admin for the required role. See Getting started.
"This workspace's API does not serve that route yet." The console asked for something the API does not offer. Reload. If it persists, tell us.
"Could not reach the Slancha API." A network failure, or a request the API refused before it could answer. Try again.
"Not available in this release." This feature is not available yet.
Signed back out to the login page. Your session ended. Sign in again. On the Admin page, a stale Google sign-in asks you to sign in with Google again.
Uploads
"remove the credentials before uploading: path:line (kind)." The package holds what looks like a secret. Remove it and upload again. The message never shows the secret.
"SKILL.md names the skill 'x'; its name here is 'y', and the two must match." Make the name in SKILL.md and the upload name agree.
"a change to 'x' is already waiting for review; approve or reject it first." A skill has one review slot.
"'x' already serves exactly this content." Nothing to propose.
"the package is not admissible: ..." The message lists what failed.
Approving
Approve is disabled. Read the caution on the card. The usual causes are below.
"this revision has not been scanned; scan it, then approve it once it is clean." Choose Scan.
"the security scan of this revision is still running." Wait. The card updates itself.
"this revision did not pass the security scan (...)." The scan found HIGH or CRITICAL issues, or its result was not clean. Read the findings. Fix the skill and upload a new version. A blocked scan cannot be overridden.
"the security scan of this revision failed; scan it again." The scan did not finish (it can also read "cancelled"). Choose Scan to run it again.
"the served revision changed since this proposal was made; propose it again." Something else was approved first. Upload the change again on top of what is served.
"the pending change was replaced after you opened it; review it again." Someone replaced the proposal while you read. Open the card again.
"acceptance criteria not met: ..." The change fails one of the skill's acceptance criteria, and the message names each failing rule. The card lists them too. Reject the change and upload a fixed version, or ask an admin to change the rules on the skill's page. Changing the rules takes effect on the waiting change at once.
Drafting
"this organization has no openai key; add one under Keys." Add the key on the Keys page, then start the job again.
"a model reads a revision only after a clean security scan." The served version of the skill has not cleared a scan. Scan it first.
"this org has no Langfuse project attached yet." An operator attaches optional tracing separately. Uploading and sharing skills does not require it.
"a new skill is written from the sessions of one project." Pick a project. If the picker is empty, no sessions were seen in the last 30 days. See Sessions.
"'x' serves variants, which drafting does not handle yet." Drafting cannot improve a skill that has a variants/ folder.
The job ends "Blocked by the security scan", "Refused by the static checks" or "The candidate was invalid". The draft was not stored. The job page gives the reason. A credential hit names the file and line.
"No traces yet." Check that the developer ran slancha auth --tracing and slancha init --tracing and restarted Claude Code, and that the Langfuse settings are in ~/.claude/settings.json.
The CLI
"No Slancha token yet." Run slancha auth, or set SLANCHA_TOKEN.
"no CLI is waiting for that code; run slancha auth again." The code expired after ten minutes, or was mistyped. Run slancha auth again.
"an org admin approves a CLI." For a tracing request, ask a workspace admin to approve the code. Members can approve their own skills-only access.
"that code was already approved." Each code is approved once.
"a proposer, approver or admin approves a CLI that asks to manage." A viewer cannot approve an agent token. Ask an admin for the proposer role, or sign in with plain slancha auth.
"This needs an agent token: run slancha auth --manage." A plain token reads published skills and nothing more. Uploading, scanning and reading reviews from the CLI need an agent token.
"Your Slancha token expired; ..." Agent tokens expire 30 days after approval. Run slancha auth --manage again. The plugin's hooks use the same token, so they stop reporting until you do.
"agent tokens cannot start drafts or outcome jobs until an admin allows it: ..." An admin turns that on in Workspace settings.
slancha upload exits 1 after printing "Uploaded". The change is stored, but its security scan did not start. Run slancha scan <skill> --wait. Do not upload it again: the skill takes no other upload while a change waits.
No tracing keys. Core slancha init works without them. For optional tracing, ask an operator to attach a project, then run slancha auth --tracing and slancha init --tracing.
Premium mining is disabled. Ask a Slancha operator to enable mining for the workspace. Core sharing remains available.
"slancha init needs: ..." Install the tools named in the message and run init again. Core setup needs claude and git; optional tracing also needs uv and a Python 3.10 or newer that uv can find.
"Claude Code already has a marketplace named langfuse-observability ... not the pinned clone." Remove that marketplace with claude plugin marketplace remove langfuse-observability, as the message says, then run init again.
"refused the download, nothing installed." A check failed. The message says which. Your installed copy is untouched.
"it has local changes." You edited an installed skill. update --force replaces it and saves your copy in .slancha-backup.
"not published yet (it is waiting for review in the console)." Approve it in the console first.
A skill does not show up in Claude Code. If ~/.claude/skills was new, run /reload-skills.