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 laurigates/claude-plugins --skill harness-tool-errorsgit clone --depth 1 https://github.com/laurigates/claude-pluginsWrote 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/laurigates/claude-plugins/harness-tool-errors)<a href="https://agentmods.dev/skills/laurigates/claude-plugins/harness-tool-errors"><img src="https://agentmods.dev/badge/skills/laurigates/claude-plugins/harness-tool-errors/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/laurigates/claude-plugins/harness-tool-errors"><img src="https://agentmods.dev/badge/skills/laurigates/claude-plugins/harness-tool-errors.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 | $0.00048 | $0.01774 |
| Opus 5.5 | $0.00019 | $0.00710 |
| Sonnet 5.5 | $0.00010 | $0.00355 |
| Haiku 4.5 | $0.00005 | $0.00177 |
Grade A, and why
harness-tool-errors 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 20d 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 — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harness Tool Errors
Promoted from the always-loaded ~/.claude/rules/tool-use-patterns.md rule,
whose stub points here. The sections below are that rule's text for the Read,
Edit/Write, WebFetch and Bash-denial failure modes.
Durable patterns distilled from weekly friction-learner reports. The
loud bash-antipatterns / branch-protection / agent-coworker-detection
hooks already enforce most of the recurring findings (git && chains, find
vs Glob, cat/head/tail, sleep-chains, etc.); the patterns below
are the failure modes those hooks don't catch.
When to Use This Skill
| Use this skill when... | Use something else when... |
|---|---|
Read errors on a missing path, EISDIR, or "exceeds maximum allowed tokens" |
An empty search result is about to gate an action → tool-result-traps |
Edit/Write refuses with "has not been read yet" or "modified since read" |
Finding every reference to a symbol before changing it → code-quality-plugin:ast-grep-search |
| A Bash call returns "Permission to use Bash has been denied" | A hook blocked the command with a message → handling-blocked-hooks.md |
| A WebFetch returns 404/403/timeout | The fetch worked but the page is empty → documentation-plugin:docs-fetch-fallbacks |
Read tool
Verify the path before calling Read
Read on a missing path is the dominant Read-side failure. The cause is
almost always an assumed cwd: the session's working directory moved
(worktree switch, prior cd, agent thread reset) and the cached path
no longer resolves.
# Wrong — three Reads against guessed paths
Read("/abs/a"); Read("/abs/b"); Read("/abs/c")
# Correct — one Glob tells you which exist
Glob(pattern="/abs/*")
Agent threads always reset cwd between Bash calls. Always pass absolute paths from an agent prompt; never assume the cwd is preserved.
Read is for files, not directories
Read on a directory errors with EISDIR: illegal operation on a directory. The error message doesn't suggest the alternative.
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.
- 20d ago Changed · +4 lines 4f38dcdd91e0
- 22d ago First seen · 168 lines · 48 tokens per session scan A 43a85755823a
harness-tool-errors is a skill published in the GitHub repository laurigates/claude-plugins (60 stars, last pushed yesterday), licensed MIT. It adds 48 tokens to every session and 1,774 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 2026-09-25.
Other skills, from other repositories
audit-dead-code
Read-only hunt for dead and unused code across four labeled lanes (Knip for TS/JS, vulture for Python, gopls for Go, grep for shell and files no lane owns), each candidate adjudicated dead, uncertain or alive. Use when asked to 'find dead code', to find what is unused in a repository (exports, functions, orphaned…
ci-log-interpretation
Use this skill when reading or analyzing CI logs from a Shopware GitHub Actions workflow to figure out why a build failed — phrases like "why did CI fail", "what broke the build", "check the pipeline", "interpret these logs", "debug this red build" — or whenever raw run logs, job logs, or check annotations from a…
devils-advocate
Stress-test plans and proposals via systematic adversarial review. Assumption extraction, evidence check, failure scenarios, operational gotchas. Before implementation begins. Use when: asked to attack a plan or proposal ('devil's advocate', 'stress test', 'poke holes', 'what could go wrong'), or before implementation…
known-issues
Lookup and tracking of known Claude product issues: GitHub bug search, service health and model quality checks, and a persistent registry of tracked issues. Use when: 'is this broken', 'any workarounds', 'troubleshoot Claude Code', 'feature behaves unexpectedly', 'scan repo for issues', 'file a bug'. Actions: status…
target
Identify and rank optimization targets by evidence quality, so an unmeasured system yields 'instrument this first' instead of a guess. Accepts session pain, a named path, telemetry, or open-ended 'what is slow here'. Use when: 'what should we optimize', 'find the bottleneck', 'this feels slow', 'is X worth…
diagnose
Diagnose unexpected behavior from Claude Code's built-in computer-use MCP server (desktop screen control). Use when: 'computer use', 'control my screen', 'screenshot is blurry', 'why is the screenshot low resolution', 'zoom in on the screen', 'screenshot capture failed', 'empty capture', 'clicks are landing in the…