dcdpr / jp / 26893797890 / 1
67%
main: 67%

DEFAULT BRANCH: main
Ran
Files 402
Run time 38s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 66.615% (+0.05%) from 66.57%
26893797890.1

push

github

web-flow 
chore(tools): Add `then` param to `debug_jp_trace` for multi-step flows (#737)

The `debug_jp_trace` tool previously ran a single `jp` command per
invocation, each in its own fresh sandbox. This made it impossible to
exercise multi-step flows — a conversation created in one call was gone
by the next, because every call re-clones from live data.

A new optional `then` parameter accepts a list of additional argument
vectors. Every command in the sequence runs inside one shared sandbox
that is created once and torn down at the end, so state created by an
earlier command is visible to later ones. For example:

```json
{"args": ["c", "new"], "then": [["q", "hello"], ["q", "and again"]]}
```

starts a conversation and then continues it twice, all in one sandbox.

On the implementation side, `execute` was split into three pieces:
`run_one` (launches one command, copies artifacts, filters events),
`execute` (wraps `run_one` for the single-command case), and
`execute_sequence` (runs `run_one` in a loop, fail-fast on error).
`trace_render` gained `render_multi` / `CommandRun` so the combined
report can use `##`-level per-command sections with `###` subsections,
while the standalone single-command report keeps its existing `##`
headings unchanged.

The skill description and MCP tool definition were updated to document
sandbox lifetime, the `then` parameter, and the tip to reach for these
tools to exercise real `jp` behavior end-to-end.

Signed-off-by: Jean Mertz <git@jeanmertz.com>

33812 of 50757 relevant lines covered (66.62%)

250.45 hits per line

Source Files on job 26893797890.1