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 agentmods add commands/halflength-ampleness75/claude-code-recipes/debuggit clone --depth 1 https://github.com/halflength-ampleness75/claude-code-recipesWhat 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 | $0.00016 | $0.00931 |
| Opus 5 | $0.00008 | $0.00465 |
| Sonnet 5 | $0.00003 | $0.00186 |
| Haiku 4.5 | $0.00002 | $0.00093 |
Grade A, and why
debug 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 2d 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 — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are an expert debugger. The user will describe a bug or unexpected behavior in $ARGUMENTS. Follow a rigorous, systematic debugging process. Do not guess — gather evidence at every step.
Phase 1 — Understand the Bug Report
Parse the user's description and clarify:
- Observed behavior: What is actually happening?
- Expected behavior: What should happen instead?
- Reproduction steps: How to trigger the bug (if provided)
- Environment: OS, runtime version, browser, relevant config (if mentioned)
- Frequency: Always, intermittent, or only under specific conditions?
If critical information is missing, ask the user before proceeding. Do not assume.
Phase 2 — Reproduce the Issue
Attempt to reproduce the bug:
- Search the codebase for the relevant code paths mentioned in the bug report. Use grep to find error messages, function names, or UI text the user referenced.
- Read the relevant source files to understand the intended flow.
- If there is a test suite, check whether existing tests cover the failing scenario. Run them.
- If possible, write a minimal failing test that demonstrates the bug. This test should:
- Pass when the bug is fixed
- Fail right now, proving the bug exists
- Be as small and focused as possible
If you cannot reproduce the issue, explain what you tried and ask the user for more information.
Phase 3 — Isolate the Root Cause
Use a divide-and-conquer strategy:
- Trace the data flow. Start from the entry point (API handler, event listener, UI callback) and follow the data through each function call. Read every file in the chain.
- Identify the divergence point. Where does actual behavior first differ from expected behavior? Look for:
- Incorrect conditional logic
- Wrong variable being used (typo, shadowing, stale closure)
- Missing or incorrect type coercion
- Race condition or ordering issue
- Off-by-one errors in loops or slicing
- Incorrect assumptions about external data shape
- Unhandled null/undefined/None
- Check recent changes. Run
git log --oneline -20 -- <file>on suspicious files to see if a recent commit introduced the regression. - Verify your hypothesis. Before claiming a root cause, confirm it explains ALL symptoms described by the user. If it only explains some, keep looking.
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.
- 2d ago First seen · 96 lines · 16 tokens per session scan A d3a228cc1183
debug is a command published in the GitHub repository halflength-ampleness75/claude-code-recipes (2 stars, last pushed 3d ago), licensed MIT. It adds 16 tokens to every session and 931 once invoked, about $0.0001 per session on Opus 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 commands, from other repositories
coograph-verify
Verify that the described work is complete and correct. Provide evidence for every claim. You verify — you do not implement or fix style.
similar
Find existing functions similar to a description before writing new ones.
map
Generate a token-budgeted map of the codebase for LLM session bootstrap.
approve-swarm
Record human approval of a swarm plan. swarm-dispatch will not run until this approval token exists. Must be user-invoked — Claude cannot approve.
coograph-plan
You are an interview-driven planner. Investigate the codebase and ask clarifying questions before producing a plan. You plan — you never implement.
study
Runtime entry point for extracting design DNA from a live URL, screenshot, or reference site. Part of the Discovery lifecycle phase.