CtrlK
BlogDocsLog inGet started
Tessl Logo

pantheon-ai/troubleshoot

Search-first troubleshooting with a diagnostic phase — use when an error, bug, or unexpected behaviour is reported.

75

Quality

75%

Does it follow best practices?

Impact

Pending

No eval scenarios have been run

SecuritybySnyk

Advisory

Suggest reviewing before use

Overview
Quality
Evals
Security
Files

troubleshoot-to-frame-problem-llm.mdreferences/

domain:
confused
verb:
decompose
constraint-type:
unknown
problem:
{problem statement}
scale:
{boulder|pebble}

Troubleshoot to Frame-Problem Handoff

Thinking Trail

  • Considered: {diagnostic approaches tried during troubleshooting}
  • Rejected: {root causes eliminated}
  • Surprised by: {problem is deeper than a simple degradation}
  • Models used: Search-first diagnostics, systematic elimination
  • Constraints discovered: {problem exceeds troubleshooting scope — needs re-framing}

Decisions

  1. Troubleshoot outcome: Stabilized but root cause requires broader change
  2. Domain shift: Complicated → Confused (re-classify for broader approach)
  3. Rationale: {fix applied but underlying issue needs architectural/process change}

Actions Taken

  • Symptom addressed: {what was stabilized}
  • Root cause identified: {deeper issue requiring re-framing}
  • Learnings saved: troubleshoot learnings database updated

Output

Stabilized: {what was fixed} Root cause: {deeper issue identified} Re-frame needed: {why broader approach required}

Domain Transition

From: Complicated (troubleshoot) → To: Confused (frame-problem) Constraint shift: governing → unknown. Troubleshooting stabilized the symptom but revealed the root cause is a different class of problem. Need fresh classification.

For /frame-problem

  • Context: Troubleshooting stabilized {symptom} but root cause is {deeper issue}
  • What's fixed: {immediate stabilization — do not re-troubleshoot}
  • What remains: {root cause requiring different approach}
  • New signals: {what troubleshooting revealed about the true problem}
  • Do NOT: Re-diagnose the symptom — it's stabilized. Classify the root cause.

Accumulated Context

Token guidance: target 300 tokens inline. For depth, use references — point to saved thinking files rather than embedding full content. Soft cap: 600 tokens inline per handoff. If you need more, move detail to a knowledge file and reference it. Accumulated cap: 800 tokens across a chain — compress to 200 at cap (keep: decisions, constraints, rejected paths). References do NOT count toward the cap.

references

reference.md

troubleshoot-to-frame-problem-llm.md

SKILL.md

tile.json