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 damianvtran/local-operator --skill systematic-debugginggit clone --depth 1 https://github.com/damianvtran/local-operatorWrote 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/damianvtran/local-operator/systematic-debugging)<a href="https://agentmods.dev/skills/damianvtran/local-operator/systematic-debugging"><img src="https://agentmods.dev/badge/skills/damianvtran/local-operator/systematic-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/damianvtran/local-operator/systematic-debugging"><img src="https://agentmods.dev/badge/skills/damianvtran/local-operator/systematic-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 | $0.00045 | $0.00553 |
| Opus 5.5 | $0.00018 | $0.00221 |
| Sonnet 5.5 | $0.00009 | $0.00111 |
| Haiku 4.5 | $0.00005 | $0.00055 |
Grade A, and why
systematic-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 5d 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Systematic debugging
Debugging is hypothesis testing against a running system, not code reading followed by a guess.
Phase 1: Reproduce
- Get a reliable, minimal reproduction: exact command, input, environment.
- Confirm you can make it fail on demand. Cannot reproduce it? Say so, collect frequency and conditions before theorising.
- Capture the failure itself: full error text, stack, exit code. Not a paraphrase.
Phase 2: Isolate
- Shrink the reproduction until the smallest failing unit is visible (scope, input size, config).
- Binary-search the surface: return early or disable at the midpoint of the suspect path. The failure is on one side of the cut; halve again.
- Instrument component boundaries. At each boundary, log what goes IN and what comes OUT. Find the FIRST boundary whose output is wrong given a correct input. The fault lives between the last correct boundary and that one.
- Let the system tell you where it breaks. You do not need to read all the code.
Phase 3: Fix the cause, or question the design
- Fixed-cause: you can say why it broke and why the fix prevents exactly that. Guessed-cause ("this might help") is not done.
- Change one variable at a time. Stacking two fixes teaches nothing about either.
- Tripwire: three or more fixes for one defect means the model is wrong or the design is. Stop patching; question the architecture or the assumption, and write down what you now know.
- Prove the fix is connected: revert it and watch the bug return, then reapply.
Phase 4: Verify and lock in
- Rerun the original reproduction; it must pass.
- Add the regression test (see
skill://test-driven-development). - Check the neighbours: the same pattern elsewhere, callers of the changed code. Evidence standards:
skill://verification-before-completion.
Red flags
- Editing before reproducing; "I think it's X" followed by a fix.
- Treating a flaky test as noise; flakiness is a real defect with real conditions.
- Suppressing the error (wider catch, retry loop, muted log) instead of finding it.
- Reading code for an hour without running anything; the running system has the answer.
- No one-sentence theory of the failure ("X fails because Y reaches it before Z").
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.
- 5d ago First seen · 43 lines · 45 tokens per session scan A ae7dea2d1c23
systematic-debugging is a skill published in the GitHub repository damianvtran/local-operator (218 stars, last pushed today), licensed MIT. It adds 45 tokens to every session and 553 once invoked, about $0.0002 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 2026-10-11.
Other skills, from other repositories
building-agents
Use when building or restructuring an LLM agent — provider adapter, tool calling, structured output, RAG, agent loop, eval gate, cost routing, tracing, MCP server — model-agnostic across OpenAI/Anthropic/Gemini/OSS so a model swap is a config change. NOT vector-store SQL alone (that is postgresdb) or service…
agent-eval
Use when measuring whether an LLM or agent system actually got better and gating merges on it: golden sets, fixing an inflated LLM-as-judge, scoring RAG (faithfulness, contextual recall) or agent trajectories (tool correctness, completion), or picking an eval framework. NOT building the agent loop, tools or RAG…
type-breakdown
Maps a code path's types, abstractions, data flow, side effects, and errors. Use only when the user explicitly asks for a type breakdown of a feature, flow, or code path.
tracelet-instrument
Instrument an AI agent or LLM app with OpenTelemetry and watch its runs live in tracelet (local DevTools). Use when asked to add tracing/observability to an agent, or to debug what an agent actually did.
hqe
Comprehensive codebase health auditing, remediation, and verification skill based on the canonical HQE Protocol v5.0.0.
systematic-debugging
Root-cause a bug with a hypothesis-driven loop instead of shotgun edits. Use when a bug isn't obvious after the first look, or when asked to "find out why" something fails.