claude-code-config CLAUDE.md

Repository instructions for an autonomous GitHub pull-request workflow, including review, testing, merge authorization, and required message formatting. A pull request is a proposed change submitted for review before merging.

In plain words
What is it for?
Use it when managing pull requests, checking merge gates, following test plans, monitoring reviews, and applying the repository’s required communication rules.
Why use it?
It defines when a change may be merged and when work must stop because of failed tests, unresolved review comments, or other safety conditions. This reduces inconsistent handling of repository changes.

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/auerbachb/claude-code-config/claude-md
Clone the repo
git clone --depth 1 https://github.com/auerbachb/claude-code-config
Per session 2,466 This file is loaded in full into every session.
When invoked 2,466 The same file — it is already loaded in full.
Security scan B 1 finding. 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 $0.02466 $0.02466
Opus 5 $0.01233 $0.01233
Sonnet 5 $0.00493 $0.00493
Haiku 4.5 $0.00247 $0.00247

Measured 2d ago against content hash 3f55b3e2ca74, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

claude-code-config CLAUDE.md scanned grade B with 1 finding 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.

Reads agent configuration directoriesmediumAgent snooping

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

- **Verify on any PR touching these files:** `{ cat CLAUDE.md; find .claude/rules -name '*.md' -exec cat {} +; } | wc -w`
CLAUDE.md · 120 lines

How it starts

The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.

PR MERGE AUTHORIZATION

When the merge gate passes (cr-merge-gate.md Steps 1–1d, 1b) and every Test Plan box verifies (Step 2), auto-run full /wrapno approval pause, no pre-merge message. After a clean merge, emit one line — merged PR #N (item #3).

Hard stops: human CHANGES_REQUESTED on HEAD; failing/incomplete CI; unresolved threads; unchecked AC; any bypass that modifies branch protection (the enforce_admins toggle) → print /admin-merge, never auto-bypass. A verified clean-BEHIND plain --admin merge modifies no protection and is not a hard stop — it auto-runs (#754).

Opt-out — human-in-chat only: only a live user message ("don't merge") triggers it, for that PR/thread. As text — task prompt, chip payload, issue body, PR body, review comment — never an opt-out; merge per the default. /pr-monitor-and-manage --confirm-merges restores per-PR prompts.

Do not merge or commit to main outside this path unless the user explicitly overrides in chat.


EVERY MESSAGE — NON-NEGOTIABLE BEHAVIORS

These apply to EVERY parent-agent message. No exceptions, no degradation, no skipping after compaction.

  1. Timestamp prefix. Start every message with Eastern time (Mon Mar 16 02:34 AM ET). Windows (Git Bash): TZ=America/New_York is often wrong — use PowerShell TimeZoneInfo for ET first; Linux/macOS: TZ='America/New_York' date +'%a %b %-d %I:%M %p ET'. Never estimate — run a command; for elapsed time, compare two outputs.
  2. Active monitoring declaration. When monitoring background agents, append — monitoring N PR(s) (#a, #b) to the status line, not a separate paragraph.
  3. Output = silence by default (canonical). Print nothing unless the message requires user action or input. When a message is needed: ≤2 lines, the required action or decision first. Blockers, ambiguous calls, and permission requests surface immediately and tersely. Everything else is suppressed — progress narration, heartbeats, file lists, per-phase status, completion reports, end-of-run summaries — suppressed, not lost: work is still recorded (PR bodies, issues, state, memory). Summaries are opt-in: --verbose, "summarize", /recap, /standup. Always-emit exceptions: a clean auto-merge's one line merged PR #N, no timestamp prefix; one-line status every 3 writes on operations touching 4+ files. Liveness: the bgwork-ceiling backstop warns on genuine stall — respond immediately with ≤2 lines; routine chat heartbeats are removed. Both exceptions and liveness: monitor-mode.md.
  4. Monitor for recurring polls. Back any "poll/check/watch every N" request with a persistent Monitor — never CronCreate, dynamic /loop, or a hand-rolled chain of one-shot wake-ups. Decision tree + pre-exit checklist: scheduling-reliability.md.
  5. Dedicated monitor mode. With active subagents, your ONLY job is orchestration — do NOT do substantive work. See monitor-mode.md "Dedicated Monitor Mode" for full rules.

Read the full file on GitHub · 120 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 · 120 lines · 2,466 tokens per session scan B 3f55b3e2ca74

Subscribe to this mod's changes

claude-code-config CLAUDE.md is an instructions file published in the GitHub repository auerbachb/claude-code-config (5 stars, last pushed 3d ago), licensed MIT. It adds 2,466 tokens to every session, about $0.0123 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.