Sessions
Optional tracing sends Claude Code sessions to your workspace's tracing project. Once enabled, it covers every session on that machine, across all repositories.
How sessions get in
Ask Slancha to attach a tracing project to your workspace. Tracing is separate from skill sharing; the console does not import transcript files.
On a developer machine, slancha auth --tracing followed by slancha init --tracing installs the tracing tool's Claude Code plugin, pinned to a fixed version, and writes your workspace's tracing keys into Claude Code's user settings. From then on Claude Code sends each turn to your workspace's tracing project. Because the keys sit in user settings, this covers every Claude Code session on that machine, in every repository.
Each turn is recorded with its prompt and its answer, and with whatever else the tracing plugin records. Image capture is turned off. Treat a session as containing anything anyone typed into Claude Code. When a session uses a skill, the turn is tagged with the skill's name, which is how Slancha finds the sessions where a skill was used.
The Slancha plugin reports skill use separately. That report holds the skill name, the version, whether the Skill tool invocation succeeded, a hash of the session id and a time. It holds no prompts and no file contents. These are Skill tool invocation counts: success comes from the hook event and tool response, not a judgment of the finished task. Slash-command invocations can have no success value. These counts do not establish task completion or that a skill improved the result.
What the console shows
The Tracing page shows tracing status: whether traces are arriving, how many since a given time (the last seven days by default), and the skills seen in them with a count each. The status reads the tracing service at most once a minute per workspace.
The Traces page lists the projects with recorded sessions and the skills seen in the window, with a Draft a skill button on each project that has enough sessions.
The New skill from traces dialog lists the projects Claude Code worked in over the last 30 days, with a session count for each.
Each skill's page shows use by version for the last 30 days: invoked, reference reads, succeeded, failed.
Coverage charts and individual-session lists are not available.
How sessions are labeled
When a draft job reads a session, it labels the session from the trace alone. No model decides the label.
| Label | Meaning |
|---|---|
| errored | A turn in the session was recorded as an error. |
| user followed up | The user said something after the agent's answer. |
| no follow-up | The session ended on the agent's first answer. |
A follow-up may be a correction, thanks or another request. It does not establish failure.
Where the outcomes show
Only sessions read by draft jobs receive labels. Find them:
- on the job's page, as counts;
- on the review card for the draft, next to each session it was written from;
- on the Tracing page, as counts beside a skill, for the sessions a draft read.
These outcomes do not grade skill quality or prove it improved a session.