An administrative debugging skill for stuck or corrupted H·AI·K·U intents, which are tracked pieces of work.

Forcing a stage to complete, changing intent fields, resetting drift, or modifying feedback after the user confirms the operation.

It provides controlled recovery operations when the normal workflow cannot continue.

haiku-debug — Skill for Claude Code and Codex by gigsmart: An administrative debugging skill for stuck or corrupted H·AI·K·U…

Install

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.

agentmods

npx agentmods add skills/gigsmart/haiku-method/haiku-debug

Any agent

npx skills add gigsmart/haiku-method --skill haiku-debug

Clone the repo

git clone --depth 1 https://github.com/gigsmart/haiku-method

Made for: Claude Code, Codex.

Or install haiku, the plugin that ships this one along with the rest of its 31 skills, 2 hooks, 2 MCP servers.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/gigsmart/haiku-method/haiku-debug.svg)](https://agentmods.dev/skills/gigsmart/haiku-method/haiku-debug)

Your own site

<a href="https://agentmods.dev/skills/gigsmart/haiku-method/haiku-debug"><img src="https://agentmods.dev/badge/skills/gigsmart/haiku-method/haiku-debug.svg" alt="Measured on agentmods" height="20"></a>

Per session 37 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,271 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

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	$0.00037	$0.01271
Opus 5	$0.00018	$0.00635
Sonnet 5	$0.00007	$0.00254
Haiku 4.5	$0.00004	$0.00127

Measured 3d ago against content hash 3c0be92b585d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

haiku-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 3d 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.

plugin/skills/haiku-debug/SKILL.md · 111 lines

How it starts

The opening of the file, as written. The whole thing — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Debug

Admin/recovery tools for intents that are corrupt, wedged, or otherwise stuck in ways the normal workflow engine can't recover from. Every mutating op routes through haiku_debug and surfaces a SPA picker for user confirmation before any state changes. The agent cannot run any admin op unilaterally.

Two surfaces — same five ops, two ways to drive them

SPA admin panel (user-driven, primary): browse to /debug on the haiku review server (default http://127.0.0.1:7320/debug). Pick an intent, see its current state + cursor preview, and run any of the five ops. The SPA's confirmation modal IS the elicitation gate — every mutation shows the exact request body before POST.
haiku_debug MCP tool (agent-driven, fallback): the agent calls the tool with op + args. The tool routes through runPicker so the user sees the same confirmation in the picker UI before any state mutates. Cancellation returns { action: "cancelled" } and no state changes.

Both surfaces call the same underlying debug-ops.ts functions. The agent cannot reach the ops without an explicit user click on either surface.

When to use this skill

An intent is stuck in a loop the engine's halt mechanism caught (loop_halted action) and the user wants to manually unblock it.
A stage's units have moved through every hat but the cursor won't advance because reviews/approvals/quality_gates aren't stamped.
The intent's mode is wrong (set to continuous when it should be autopilot, etc.) — mode is normally engine-managed, but mid-flight changes need an override.
Drift sweep keeps re-firing on stale witnesses even after the underlying state matches — re-stamp every witness to the current SHA.
A feedback record is in an impossible state (closed but resolution still null, etc.) and needs surgical FM mutation.

When NOT to use this skill

Use /haiku:haiku-repair first for the supported v4 recovery paths (drift baseline, worktree relocation, mainline PR generation).
Use /haiku:haiku-reset-stage or /haiku:haiku-reset-intent for full destructive resets.
For day-to-day workflow questions (why didn't this stage advance? what's the cursor doing?), reading the prompt file or haiku_intent_list is usually faster.

Read the full file on GitHub · 111 lines

Changes

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.

3d ago First seen · 111 lines · 37 tokens per session scan A 3c0be92b585d

Subscribe to this mod's changes

haiku-debug is a skill published in the GitHub repository gigsmart/haiku-method (24 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 37 tokens to every session and 1,271 once invoked, about $0.0002 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 2026-08-30.

Debugging ai ai-agents ai-dlc claude development-lifecycle methodology

mass-line

触发：当你需要收集多方意见、把零散反馈整合成可执行方案，或把方案带回真实使用者/执行者验证时调用；常见信号包括 stakeholder input、user feedback、意见汇总、对齐与验证。 English: Trigger when input must be gathered from many people, synthesized into a clearer plan, and returned to the affected users or executors for validation. Use this skill for a collect-synthesize-validate loop.

HughYau/qiushi-skill · 103 tokens

workflows

触发：当你面临的任务明显需要多个思想武器协作时调用；常见信号包括：从零启动新项目、攻坚复杂疑难问题、对已有方案进行迭代优化。此 skill 提供标准化的跨 skill 工作流组合，解决"应该先用哪个 skill、怎么衔接"的问题。 English: Trigger when a task clearly requires multiple skills in sequence. Use this skill to select a standard workflow that chains skills together, defines data handoff between steps, and specifies…

HughYau/qiushi-skill · 117 tokens

concentrate-forces

触发：当多个任务同时争夺时间、注意力、算力或预算，必须确定主攻方向并停止分散用力时调用；常见信号包括优先级过多、资源紧张、推进分散、需要决定先做什么。 English: Trigger when limited resources are being split across too many tasks and one main target must be chosen. Use this skill to concentrate effort, sequence work decisively, and finish a meaningful breakthrough before expanding.

HughYau/qiushi-skill · 105 tokens

practice-cognition

触发：当你提出了方案、假设或判断，需要通过实践验证、试错迭代或复盘升级认知时调用；常见信号包括 experiment、prototype、validate、iterate、feedback loop。 English: Trigger when an idea, hypothesis, or plan must be tested in practice and improved through iteration. Use this skill to move from action to understanding and back to action in a spiral learning loop.

HughYau/qiushi-skill · 92 tokens

contradiction-analysis

触发：当问题复杂、存在多个冲突因素、优先级不清，或你不知道应该先解决什么时调用；常见信号包括 trade-off、瓶颈、根因不明、主次不清、多个问题互相牵制。 English: Trigger when a problem contains competing forces, unclear priorities, or no obvious entry point. Use this skill to identify contradictions, isolate the principal contradiction, classify its nature, and choose the right response.