Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add stuartshields/claude-setup --skill debugginggit clone --depth 1 https://github.com/stuartshields/claude-setupWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/stuartshields/claude-setup/debugging)<a href="https://agentmods.dev/skills/stuartshields/claude-setup/debugging"><img src="https://agentmods.dev/badge/skills/stuartshields/claude-setup/debugging/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/stuartshields/claude-setup/debugging"><img src="https://agentmods.dev/badge/skills/stuartshields/claude-setup/debugging.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00071 | $0.00609 |
| Opus 5.5 | $0.00028 | $0.00244 |
| Sonnet 5.5 | $0.00014 | $0.00122 |
| Haiku 4.5 | $0.00007 | $0.00061 |
Grade A, and why
debugging scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 32 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debugging & Investigation
4-Step Framework: Reproduce → Isolate → Fix → Validate
- Never skip to Fix. Most wrong fixes come from jumping straight to code changes.
- Reproduce: Run the faulty scenario. Note the exact error and conditions.
- Isolate: Narrow down the failing component. 80% of bugs live in recently modified code — start there. Compare against working code. Find similar functionality that works and diff against the broken code.
- Validate: Run the full test suite. Confirm the error is gone AND no regressions introduced.
Validate Before Fixing
- Confirm root cause before proposing code changes. One hypothesis at a time — test the most likely first.
- Explain root cause in writing (file path, line number, actual vs expected behavior) before proposing code changes.
- Test means run code, not reason in your head. Execute a command, write a failing test, or check real output.
- Read the user's evidence first. Study screenshots/errors before forming hypotheses. The answer is usually visible.
- Add
console.errorto trace actual behavior when reasoning about what code "should" do isn't working. Reading code is not observing behavior. - Red flag phrases. Catch yourself saying "This should work", "Let me just try...", or "Quick fix for now" — stop and re-evaluate. You're guessing.
Anti-Loop Protocol
- Each attempt MUST use a different diagnosis. If attempt 1 failed, don't vary the fix — vary the diagnosis. Two attempts targeting the same component with the same assumption count as one diagnosis, even if the code changes differ.
- After 2 failed fix attempts, STOP. State what you tried, what each ruled out, and ask the user for context. If the user says "still broken" twice, you've failed twice.
- Clear polluted context. After hitting the limit, recommend
/clear— failed attempts in context actively degrade subsequent reasoning. - Watch for oscillation. Fixing A breaks B and vice versa = contradictory constraints. Stop and surface the conflict.
- Revert as escape hatch. When debugging spirals, revert to last known good commit rather than stacking fixes on failed attempts.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 32 lines · 71 tokens per session scan A 2edcefe77471
debugging is a skill published in the GitHub repository stuartshields/claude-setup (2 stars, last pushed 4mo ago), licensed MIT. It adds 71 tokens to every session and 609 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
debug
Instrument web/web-app code with structured debug logging via a global variable (window.debuglogs). Produces a clean JSON timeline for reproducing and diagnosing bugs. Use when user wants to debug a feature or track down a bug.
transcript-grep
Search a Claude Code session transcript for a string or regex pattern and show every matching message with surrounding context. Reads /api/sessions/:id/transcript and resolves sessions via /api/sessions?limit= from the Agent Monitor dashboard. Use when hunting for a specific message, prompt, tool call, or error inside…
anomaly-alert
Identify anomalous sessions using Agent Monitor data — cost outliers from the pricing engine, token anomalies (cache miss spikes, compaction baseline surges), unusual event type ratios (PreToolUse/PostToolUse gaps, APIError clusters), behavioral deviations from workflow intelligence (complexity score outliers, error…
regression-watch
Detect quality and efficiency regressions over time using Agent Monitor data — rising error rate (APIError events), falling cache hit rate, growing compaction frequency, and climbing cost-per-session. Splits history into an earlier baseline window and a recent window and reports which metrics are getting worse, by how…
hook-inventory
Inventory hooks across the user, project, and project-local settings plus the /.claude/hooks scripts directory — read through the Agent Monitor Config Explorer API — and flag hooks that POST to the network or run arbitrary commands. Reads /api/cc-config/hooks and /api/cc-config/hook-scripts. Use when auditing hook…
event-trace
Trace the full event chain for one Claude Code session into an ordered timeline of every event type with toolname and summary, highlighting gaps, out-of-order events, and failures. Reads /api/events?sessionid= and /api/sessions/:id from the Agent Monitor dashboard. Use when debugging what a session actually did, step…