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 chan4lk/specclaw --skill debuggit clone --depth 1 https://github.com/chan4lk/specclawWrote 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/chan4lk/specclaw/debug)<a href="https://agentmods.dev/skills/chan4lk/specclaw/debug"><img src="https://agentmods.dev/badge/skills/chan4lk/specclaw/debug/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/chan4lk/specclaw/debug"><img src="https://agentmods.dev/badge/skills/chan4lk/specclaw/debug.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 | $0.00024 | $0.01889 |
| Opus 5.5 | $0.00010 | $0.00756 |
| Sonnet 5.5 | $0.00005 | $0.00378 |
| Haiku 4.5 | $0.00002 | $0.00189 |
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 4d 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 — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
specclaw debug
First, run specclaw-ensure-init .specclaw — idempotently creates .specclaw/ if it doesn't exist (silent if already initialized; auto-inits using the current directory's basename as the project name).
The Iron Law
No fix without a stated root cause.
One hypothesis at a time. The smallest change that tests that hypothesis. A wrong hypothesis is withdrawn and replaced — never fixed on top of.
The cost of breaking this law is not a wasted turn. It is a codebase carrying three speculative changes, none of which addressed the problem, and no record of what was ruled out — so the next person starts from zero.
Phase 1 — Evidence
Before forming any theory:
- Read the whole error. Not the first line. Stack traces name the boundary that failed; the first line names the boundary that noticed.
- Reproduce it with one command. Write that command down — it is the record's
reproducefield and the thing the fix must be proved against. If it cannot be reproduced on demand, say so explicitly; an unreproducible bug is a different investigation. - Find when it last worked.
git log,git diffsince that point. A bug that appeared in a known window has a much smaller search space than one that did not. - In multi-component paths, instrument the boundaries before theorising. Print or log what crosses each seam. A guess about which side is wrong costs more than the two minutes of instrumenting that answers it.
Observations only in this phase. "The config is probably not loading" is not an observation.
Phase 2 — Compare
Find the nearest thing in this repo that works — the sibling test that passes, the other caller of the same function, the module that does the same job correctly — and list every difference between it and the failing case. Most root causes are in that list.
Phase 3 — One hypothesis
Write it down, in this form:
Hypothesis: <mechanism> because <evidence>
because is not optional. A hypothesis with no evidence behind it is a guess, and guesses are what
this protocol exists to stop.
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.
- 4d ago Changed 609fec873b05
- 28d ago First seen · 153 lines · 24 tokens per session scan A 6b155f270f4d
debug is a skill published in the GitHub repository chan4lk/specclaw (13 stars, last pushed today), licensed MIT. It adds 24 tokens to every session and 1,889 once invoked, about $0.0001 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 2026-09-19.
Other skills, from other repositories
loki-mode
Launch Loki Mode autonomous SDLC agent. Handles PRD-to-deployment with minimal human intervention. Invoke for multi-phase development tasks, bug fixing campaigns, or full product builds.
reground
The current session has drifted: the recent work rests on assumptions the codebase has not been shown to support, invents systems that may not exist, or builds past what the task needs. This skill is a hard reset to ground truth. It does not summarize or compact — it discards the speculative thread and rebuilds a…
runbook
Investigate production issues from symptom to structured report. Use this skill when something is broken, slow, or behaving unexpectedly in production or staging. Triggers on "it's broken", "users are reporting", "alert fired", "error in prod", "why is X slow", "investigate", "incident", "outage", "debug production"…
remote-fix
Remote fix workflow. Receives a diagnostic (from Cowork scheduled task, manual observation, or user report), investigates the issue, fixes it, and creates a PR. Designed for quick fixes triggered remotely via Channels (Discord/Telegram). Use when the user says "fix this", "handle this diagnostic", "the scheduled task…
diagnose-bug
A bug-investigation workflow that reproduces a reported problem, identifies its cause, and writes a task for implementation. It can also prepare a planning brief when the fix changes an API contract, database structure, dependency, or project settings.
systematic-debugging
4-phase root cause debugging: understand bugs before fixing.