codexbar CLAUDE.md

Project instructions for Codexbar, a small Bash-based desktop status tool with no build step or external dependencies.

In plain words
What is it for?
Use them when changing the Bash script, its desktop panel, installation behavior, fallback logic, or related tests.
Why use it?
They document edge cases around command detection, bundled fallbacks, URL decoding, and the fixed data format that connects its parts.

Instructions file

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 instructions/mryll/codexbar/claude-md
Clone the repo
git clone --depth 1 https://github.com/mryll/codexbar
Per session 849 This file is loaded in full into every session.
When invoked 849 The same file — it is already loaded in full.
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.

ModelPer sessionOnce invoked
Fable 5 $0.00849 $0.00849
Opus 5 $0.00425 $0.00425
Sonnet 5 $0.00170 $0.00170
Haiku 4.5 $0.00085 $0.00085

Measured 2d ago against content hash b5e25a8600f9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codexbar CLAUDE.md 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.

CLAUDE.md · 33 lines

How it starts

The opening of the file, as written. The whole thing — 33 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Tooling

  • Single bash script (codexbar) — no build step, no dependencies to install
  • Install: make install PREFIX=~/.local

Non-Obvious Rules

  • Quickshell emits NEITHER started NOR exited when the command does not existrunning just drops back to false. sawExit is the discriminator: no exited = the run could not start; an exited run with empty output is an operational failure, never "not installed". Probed live: exited always arrives before running drops.

  • The bundled-script fallback fires ONLY on a failed start. The plugin clone carries the script, so the panel switches resolvedBin to the clone's copy when PATH cannot start it. PATH first, always: the AUR release wins when it exists. Never fall back on operational errors. A later yay -S applies on the next shell restart.

  • URL→path decodes each segment (mirror of Util.fileUrl). The naive scheme strip is banned; test_bundled_fallback.sh pins it. A bad % degrades to PATH-only.

  • schema_version is pinned to 2 on both sides. A schema bump must change script, panel and tests in one commit.

  • installCmd is the one constant — the message shows it and the button copies it (Util.execArgv(["wl-copy", ...]), no shell line, no trailing newline). The button gates on notInstalled, never on error text.

  • A tooltip meter is PARKED, not rendered in place. The bar has to reach the tooltip's right edge, and that edge is the widest TEXT line — which does not exist yet while the lines are being collected. So a meter pushes METER:<i> into lines plus one entry in the parallel meter_* arrays, and the width pass resolves it. The width pass MUST skip METER: lines, or the measurement is circular. Every meter in one tooltip gets the SAME bar length: they stack, so a reader compares them against each other.

  • The script must ALWAYS exit 0 and output valid Waybar JSON ({text, tooltip, class}), even on errors — use the die() helper for error paths

  • Tooltip markup is Pango (not HTML) — Waybar renders it via GTK/Pango

  • Bar text is also wrapped in Pango <span> for coloring

  • set -euo pipefail is active — unset variables and failed pipes are fatal

  • Cache lives at ~/.cache/codexbar/; auth at ~/.codex/auth.json

  • Concurrent instances are serialized with flock (multi-monitor support)

  • Color chain (first match wins): --color-* flags > Omarchy theme (~/.local/state/omarchy/current/theme/colors.toml, legacy ~/.config/...) > pywal (${XDG_CACHE_HOME:-$HOME/.cache}/wal/colors.json) > One Dark built-ins. Every step degrades silently; the resolved anchors are published as palette in --json so the Quickshell panel draws the same gauge

  • IFS=$'\t' read collapses runs of empty fields (TAB is IFS whitespace) — emit a sentinel from jq, never an empty TSV field

Read the full file on GitHub · 33 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.

  1. 2d ago First seen · 33 lines · 849 tokens per session scan A b5e25a8600f9

Subscribe to this mod's changes

codexbar CLAUDE.md is an instructions file published in the GitHub repository mryll/codexbar (14 stars, last pushed 5d ago), licensed MIT. It adds 849 tokens to every session, about $0.0042 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.