Debug escalation

Skill tarangdeep-goel-by/claude-harness/claude/skills/debug-escalation

Dev-focused Claude Code harness — workflow-engine skills, in-repo memory, session continuity (/recall ↔ /vault-push), local telemetry, + a ready-to-fill vault scaffold. clone + ./install.sh and go.

Install

npx -y skills add tarangdeep-goel-by/claude-harness --skill debug-escalation

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when encountering any bug, test failure, unexpected behavior, or error during development, before proposing or attempting fixes.

SKILL.md

3.1 KB, as published. Nobody here has run it

Debugging with Escalation

Process

Phase 1: Understand Before Fixing

Do NOT propose a fix until you have:

Read the complete error message (full stack trace, exit code, line numbers)
Identified the failing line and the actual vs expected values
Traced backward through the call stack to find the origin of the bad value
Checked recent changes (git diff) that might have caused it

If the user reports an error: Ask for specific input/output before guessing.

Phase 2: Single Hypothesis

Form ONE hypothesis: "I think X happens because Y"
Test with the SMALLEST possible change
Verify: did it work? Check the actual output.
If it didn't work, form a NEW hypothesis — do NOT layer another fix on top

Phase 3: Escalation

Track your fix attempts. After each failed attempt, note what you tried and why it failed.

After 3 failed fix attempts: STOP.

This means your mental model is wrong. Do not attempt a 4th fix. Instead:

Summarize what you've tried and why each failed
State what this tells you about the actual problem
Ask: "This might be an architectural issue rather than a simple fix. Here's what I've found — how do you want to proceed?"

Escalation Signals from User

If the user says any of these, change your approach IMMEDIATELY:

"Stop guessing"
"That's not it"
"Try something different"
"Why does this keep happening?"
Any sign of frustration with repeated attempts

When you hear these: stop, summarize findings, ask for direction.

Logging (MANDATORY)

When invoked, and on escalation (3rd failure), log via Bash:

echo '{"ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","skill":"debug-escalation","project":"<project>","attempt":<1|2|3>,"escalated":<true|false>,"task":"<brief description>"}' >> ~/vault/logs/workflow.jsonl

Example Walkthrough

Error: TypeError: Cannot read properties of undefined (reading 'map') at Dashboard.tsx:42

Phase 1 (Understand):

Line 42: data.items.map(...) — data is defined but data.items is undefined
Check the API response shape — the endpoint returns { results: [...] }, not { items: [...] }
git diff shows the API was recently changed from items to results

Phase 2 (Hypothesis):

Hypothesis: "The frontend still expects items but the API now returns results"
Smallest fix: change data.items to data.results on line 42
Run the app → works

If instead the fix didn't work, form a NEW hypothesis — don't add a fallback on top.

Anti-Patterns

Pattern	Fix
Stacking fixes without verifying each	One change at a time, verify between each
"Let me try one more thing" after 3 failures	STOP. Escalate.
Guessing without reading the error	Read the full error message first
Changing multiple things at once	Isolate: one variable at a time
Assuming the fix worked without running it	Run the command. Read the output.