weft-debug
SkillDev toolsLets your agent debug the last run of a Weft workflow, tracing the failure to one node and proposing a fix.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the weft-debug skill
About this capability
COMMAND, not reference: Inspect the last execution and chase the failure to one node. Run this when the user asks for this step by name, optionally naming an execution color. The sixteen `weft-` reference skills beside it are things you read; this is a procedure you carry out.
What this skill tells your AI
The instructions your AI receives, as published by weavemindai/weft in tangle/cline/.cline/skills/weft-debug/SKILL.md and read by ahel’s review.
Debug this project's latest run, or a specific execution color if the user named one. Read the weft-running skill if you have not this session.
weft executions --limit 5 --phase fireand pick the color: the one the user named, else the latest.weft logs <color>first: it names the failing node and quotes the error, and for most failures that is the whole diagnosis. Then narrow the events instead of reading them all:weft events <color> --kind failedfor the failures,--node <id>for what reached one node,--kind node_skippedfor what did not run and why,--fullon the one line whose value you need whole.- Chase it to one node:
- A failed node: name it, quote its error, list the values that reached its inputs (its
node_startedline), and read the relevant source. - A wrong value: the motion backwards. Look at the top-level groups in
src/main.weft, find the first whose output is already wrong in the events, descend into it, repeat, until you hold the single step whose value went wrong and the concrete input that broke it. - Suspended: say where it waits and for whom; if nobody should have been asked, the bug is the
_should_flowthat routed there. - Cancelled: read the reason on its
execution_cancelledline.Stopped by execution <color> (tag <tag>)means another run asked the engine to stop everything carrying that tag; look at that run, and if no sibling was meant to, the bug is the tagging. Theweft-runningskill reads every cancel reason. - Nothing ran: nothing reached the branch; walk from the trigger (or the root) down to the first node whose
_should_flowor required input closed.
- A failed node: name it, quote its error, list the values that reached its inputs (its
- Name one fix, in two or three sentences: what you will change and why that fixes the observed value, not just the symptom. Apply it, validate, and run the same case again. Ask first only when the fix changes what the program does for the user, or when it would spend money or destroy state.
Report, in plain words: the color and status, the node and value at fault, the fix and whether it ran again green or waits on the user's yes. No hedging, no blaming the environment.
Signals
- GitHub stars
- 2k
- Forks
- 221
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
weft-debug- Source
- github.com/weavemindai/weft