mode-debug
SkillDev toolsGuides your agent through a disciplined observe-hypothesize-test loop to find and fix bugs.
Use mode-debug in Claude, ChatGPT or Ahel Desktop
Free. Sign in, add mode-debug and connect your AI. About a minute.
Also: Claude Code · Cursor · Codex
Then ask your AI: use the mode-debug skill
Details
Instructions available. Your AI can read the instructions. Execution depends on the setup they require.
Account requirements not reviewed. Check the skill instructions before use; Ahel provides instructions and does not run this skill.
No other account needed.
Add Ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.
About this skill
Debug mode: observe, hypothesize, test, evaluate, fix-and-stop; no certainty without evidence, no symptom patches. Load when switching to [debug mode] or when diagnosing a bug or unexpected behavior.
What this skill tells your AI
The instructions your AI receives, as published by weavemindai/weft in .agents/skills/mode-debug/SKILL.md and read by Ahel’s review.
Diagnose and fix bugs. Follow the loop strictly:
- Observe. Read the relevant code, trace the execution path. Facts before opinions.
- Hypothesize. State one hypothesis: "maybe the issue is X". Never claim certainty without evidence; never say "the issue is X" out of the blue, observation and hypothesis come first.
- Test. Design a test whose result confirms or rejects the hypothesis, and know in advance which result does which. When you add debug logging, overdo it: one run with full information beats ten recompile-retest cycles. If you can't run the probe yourself, ask the [user] to run it and paste the output.
- Evaluate. Confirmed: fix. Rejected: new hypothesis, back to step 2. Never patch the symptom. Stuck in a loop: stop and ask the [user]'s input.
- Fix and stop. Implement the fix, then stop and ask the [user] to test. No additional changes before verification.
If you catch yourself writing "let me also..." or "but there's still an issue" after implementing a fix, write verbatim "Wait stop. I already implemented a fix that should cover that let me at least test it before suggesting another fix on top." and test instead or ask the user to test if you can't yourself.
If something fails and you don't know why: switch to [research] mode and search online. You do not guess. You do not suggest workarounds.
A doubt encountered here is a task, never a shrug: any "huh, that's strange" during debugging gets investigated to the bottom, right then.
Signals
- GitHub stars
- 2k
- Forks
- 219
- Last commit
- Oct 2026
Advanced
- Item type
- skill
- Key
mode-debug- Source
- github.com/weavemindai/weft