claude-harness-canary

claude-harness-canary is a skill for Claude Code, Codex from gtapps/claude-code-hermit. It costs 74 tokens per session (4,418 once invoked), scanned D, original, MIT.

A compatibility check for Claude Code, an interactive coding assistant, run through real terminal sessions before a core release or after Claude Code changes.

In plain words
What is it for?
Use it to verify the current Claude Code setup through interactive tmux sessions and collect evidence that the supported workflows still behave correctly.
Why use it?
It catches problems with model switching, unattended prompts, and built-in monitoring or scheduled-task routines before they reach users.

Skill for Claude CodeCodex

Written for Claude Code and Codex: SessionStart hook event, but also agents/openai.yaml present. Also seen: reads .claude/ paths; names the AskUserQuestion tool; mentions Claude Code.

Good fit Use it to verify the current Claude Code setup through interactive tmux sessions and collect evidence that the supported workflows still behave correctly.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/gtapps/claude-code-hermit/claude-harness-canary
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.

Any agent
npx skills add gtapps/claude-code-hermit --skill claude-harness-canary
Clone the repo
git clone --depth 1 https://github.com/gtapps/claude-code-hermit

Made for: Claude Code, Codex.

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.

agentmods badge for claude-harness-canary

README.md
[![agentmods](https://agentmods.dev/badge/skills/gtapps/claude-code-hermit/claude-harness-canary/github.svg)](https://agentmods.dev/skills/gtapps/claude-code-hermit/claude-harness-canary)
Your own site
<a href="https://agentmods.dev/skills/gtapps/claude-code-hermit/claude-harness-canary"><img src="https://agentmods.dev/badge/skills/gtapps/claude-code-hermit/claude-harness-canary/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.

agentmods 80×15 button for claude-harness-canary

Your own site · 80×15
<a href="https://agentmods.dev/skills/gtapps/claude-code-hermit/claude-harness-canary"><img src="https://agentmods.dev/badge/skills/gtapps/claude-code-hermit/claude-harness-canary.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 74 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,418 The whole file, excluding the scripts and references it only reads on demand.
Security scan D 2 findings. A grade says what 26 rules found in the file — not that it is safe.
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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00074 $0.04418
Opus 5 $0.00037 $0.02209
Sonnet 5 $0.00015 $0.00884
Haiku 4.5 $0.00007 $0.00442

Measured 5d ago against content hash 89e46a83a64f, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade D, and why

claude-harness-canary scanned grade D with 2 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

"<project>/.claude/settings.local.json" permissions-sync

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

Never use `rm -rf`.
.agents/skills/claude-harness-canary/SKILL.md · 386 lines

How it starts

The opening of the file, as written. The whole thing — 386 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Claude Harness Canary

Run all three groups serially against the currently installed Claude Code. Take no arguments and do not offer group selectors. Target 5–8 minutes, but preserve bounded waits and evidence over that estimate.

Keep this canary core-only, current-Claude-only, interactive-tmux-only, and loaded through --plugin-dir. Never add it to CI, release commands, or $pre-release-review.

Shared probe prerequisite

Reuse the installed $probe lifecycle. Before creating canary state, read its active SKILL.md and require all of these properties:

  • real interactive Claude in tmux, never claude -p, Codex, or another PTY;
  • no interactive --no-session-persistence;
  • tmux pipe-pane starts before any input;
  • one bounded readiness phase that detects the trust dialog, accepts it only when present, handles only explicitly recognized startup onboarding, and waits for the normal prompt;
  • a unique output-only completion nonce whose complete value is never contiguous in the injected prompt;
  • completion accepted only from assistant output after the requested evidence, never prompt echo;
  • one bounded wait per canary action;
  • teardown with rm -r only after validating the exact scratch path.

If any property is absent, stop before preflight and report:

PREREQUISITE BLOCKED — update the local $probe protocol

Do not recreate generic tmux lifecycle rules locally or hand-roll a fallback. Extend the probe lifecycle only for the multi-action assertions below.

Preflight

Run from the repository root.

  1. Resolve the absolute core path at plugins/claude-code-hermit.

  2. Record bounded output from:

    claude --version
    claude --help
    tmux -V
    claude auth status --json
    
  3. Require current help to expose --plugin-dir, --model, --permission-mode, and --session-id.

  4. Require authentication through the default profile. Do not set CLAUDE_CONFIG_DIR.

  5. Require these core files:

    • .claude-plugin/plugin.json
    • hooks/hooks.json
    • scripts/lib/harness-command.ts
    • scripts/stop-pipeline.ts
    • scripts/hermit-watchdog.ts
    • scripts/ask-gate.ts
    • scripts/apply-settings.ts
    • scripts/routines.ts
    • scripts/routine-monitor.sh
    • skills/hermit-routines/SKILL.md

Read the full file on GitHub · 386 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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.

  1. 5d ago Changed 89e46a83a64f
  2. 9d ago First seen · 386 lines · 74 tokens per session scan D 046520074ec0

Subscribe to this mod's changes

claude-harness-canary is a skill published in the GitHub repository gtapps/claude-code-hermit (74 stars, last pushed today), licensed MIT. It adds 74 tokens to every session and 4,418 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it D with 2 findings (reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.